219 128 4MB
English Pages 560 Year 2013
THE ECONOMIC ASSESSMENT OF MERGERS UNDER EUROPEAN COMPETITION LAW
This concise and practical guide to the most important economic techniques and evidence employed in modern merger control draws on the authors’ extensive experience in advising on European merger cases. It offers an introduction to the relevant economic concepts and analytical tools, and stand-alone chapters provide an in-depth overview of the theoretical and practical issues related to market definition, unilateral effects, coordinated effects and non-horizontal mergers. Each form of economic evidence and analysis is illustrated with practical examples and an overview of key merger decisions. daniel gore is a principal with RBB Economics, a consultancy specialising in the economics of competition policy. stephen lewis is a principal with RBB Economics. andrea lofaro is a partner at RBB Economics. frances dethmers is a counsel economist with leading international law firm Clifford Chance LLP.
THE ECONOMIC ASSESSMENT OF MERGERS UNDER EUROPEAN COMPETITION LAW DANIEL GORE, STEPHEN LEWIS, ANDREA LOFARO and FRANCES DETHMERS
c a m b r i d g e u n i v e r s i t y p re s s Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo, Delhi, Mexico City Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9781107007727 © Frances Dethmers, Daniel Gore, Stephen Lewis and Andrea Lofaro 2013 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2013 Printed and bound in the United Kingdom by the MPG Books Group A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Gore, Daniel. The economic assessment of mergers under European competition law / Daniel Gore, Stephen Lewis, Andrea Lofaro and Frances Dethmers. pages cm ISBN 978-1-107-00772-7 (Hardback) 1. Consolidation and merger of corporations–Europe. 2. Competition–Law and legislation–Europe. I. Title. HD2746.5.G667 2013 338.80 3094–dc23 2012046195 ISBN 978-1-107-00772-7 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.
CONTENTS
Table of Merger Decisions viii Acknowledgments xiii Foreword by Nicholas Levy xv A Comment on the Text by Professor Jeffrey Church 1
Introduction
1
1. Purpose and scope of this book 1 2. Legal framework and Commission procedure 2.1 The Merger Regulation 3 2.2 Procedure for notification and assessment 3. The role of economics in EU merger control 4. Basic economic concepts 13 4.1 The demand curve 14 4.2 The cost curve 20 23 4.3 Profit maximisation 4.4 Market power 24 5. Empirical economic evidence 27
2
Market definition
xvii
3 7 12
31
1. Introduction 31 2. Conceptual framework 33 2.1 The hypothetical monopolist/SSNIP test 33 2.2 Issues and complications arising in market definition 3. Empirical techniques to assess market definition 43 3.1 Critical loss analysis 46 3.2 Demand estimation 64 3.3 Survey evidence 74 3.4 Analysis of sales patterns 81 3.5 Analysis of price levels 93 3.6 Price correlation analysis 105 3.7 Stationarity analysis 127 3.8 Shock analysis 134
v
36
vi
contents
3
Horizontal mergers I: unilateral effects
148
1. Introduction 148 2. Important competitive constraints: economic theory and the EC 150 Horizontal Merger Guidelines 2.1 Economic theory and key concepts 150 2.2 Important competitive constraints in the EC Horizontal Merger Guidelines 158 163 2.3 Commission’s recent enforcement practice 3. Important competitive constraints between the parties: 176 empirical techniques 3.1 Survey evidence 177 3.2 Customer switching analysis 183 3.3 Price/concentration studies and analysis of the impact of rivals’ presence 191 3.4 Entry analysis 207 3.5 Natural experiments 211 3.6 Merger simulation 214 3.7 Win/loss and bidding analysis 231 4. Further issues relevant in the assessment of unilateral effects 4.1 Introduction 251 4.2 Elimination of potential competition 252 4.3 Rivals’ ability to increase supply 258 4.4 Switching costs 275 4.5 Partial ownership 282 5. Countervailing factors 288 5.1 Introduction 288 5.2 Buyer power 288 5.3 Product repositioning and new entry 298 5.4 Efficiency analysis 304 5.5 Failing firm defence 310
4
Horizontal mergers II: coordinated effects
251
318
1. Introduction 318 2. Economic concepts 321 2.1 Textbook tacit coordination 321 2.2 Factors that affect the critical discount factor 326 3. Framework for the assessment of coordinated effects 335 3.1 Will the merged entity and remaining competitors be able to reach a tacit understanding? 337 3.2 Are market characteristics such that any tacit understanding 349 would likely be sustained? 3.3 Will the proposed transaction make it significantly more likely that tacit coordination will occur or make tacit coordination more effective? 368
vii
contents
5
Non-horizontal mergers
378
1. Introduction 378 2. Economic concepts 380 2.1 Pro-competitive effects 380 2.2 Foreclosure and anti-competitive foreclosure 385 2.3 Other anti-competitive effects 389 3. Vertical mergers: input foreclosure 391 3.1 Ability to engage in input foreclosure 393 3.2 Incentive to engage in input foreclosure 409 3.3 Effect of input foreclosure 428 4. Vertical mergers: customer foreclosure 430 4.1 Ability to engage in customer foreclosure 432 4.2 Incentive to engage in customer foreclosure 439 4.3 Effect of customer foreclosure 446 4.4 Customer foreclosure: EU case law 448 5. Conglomerate mergers 452 5.1 Ability to foreclose in conglomerate mergers 455 5.2 Incentive to foreclose in conglomerate mergers 458 5.3 Effects of foreclosure in conglomerate mergers 461 5.4 Foreclosure in conglomerate mergers: an example 464 5.5 Assessment of foreclosure in conglomerate mergers: EU case law 480 6. Diagonal mergers 502 6.1 Diagonal mergers: an example 503 6.2 Case study: Google/Doubleclick – assessment of diagonal effects 504
Appendix A Regression analysis and econometrics
507
1. Regression analysis and statistical inference 507 2. Regression analysis, endogeneity and the identification of economic effects 515
Appendix B
Models for demand estimation
520
1. Introduction 520 1.1 Alternative models for continuous demand estimation 1.2 Models for discrete demand estimation 522
Index
526
520
TABLE OF MERGER DECISIONS
Case COMP/M.4980 - ABF/GBI Business, Commission decision of 23 September 2008. Case COMP/M.4647 - AEE/Lentjes, Commission decision of 5 December 2007. Case COMP/M.053 - Aerospatiale-Alenia/de Havilland, Commission decision of 2 October 1991. Case COMP/M.1630 - Air Liquide/BOC, Commission decision of 18 January 2000. Case COMP/M.3314 - Air Liquide/Messer Targets, Commission decision of 15 March 2004. Case COMP/M.1524 - Airtours/First Choice, Commission decision of 22 September 1999. Case COMP/M.1681 - AKZO Nobel/Hoechst Roussel Vet, Commission decision of 22 November 1999. Case COMP/M.1715 - Alcan/Pechiney, withdrawn on 14 March 2000. Case COMP/M.1601 - Allied Signal/Honeywell, Commission decision of 1 December 1999. Case COMP/M.3099 - Areva/Urenco/ETC JV, Commission decision of 6 October 2004. Case COMP/M.4513 - Arjowiggins/M-real Zanders Reflex, Commission decision of 4 June 2008. Case COMP/M.5153 - Arsenal/DSP, Commission decision of 9 January 2009. Case COMP/M.2314 - BASF/Eurodiol/Pantochim, Commission decision of 11 July 2001. Case COMP/M.993 - Bertelsmann/Kirch/Premiere, Commission decision of 27 May 1998. Case COMP/M.3178 - Bertelsmann/Springer/JV, Commission decision of 3 May 2005. Case COMP/M.3625 - Blackstone/Acetex, Commission decision of 13 July 2005. Case COMP/M.1879 - Boeing/Hughes, Commission decision of 29 October 2000. Case COMP/M.877 - Boeing/McDonnell Douglas, Commission decision of 30 July 1997. Case COMP/M.2139 - Bombardier/ADtranz, Commission decision of 3 April 2001. Case COMP/M.3975 - Cargill/Degussa Food Ingredients, Commission decision of 29 March 2006. Case COMP/M.6106 - Caterpillar/MWM, Commission decision of 19 October 2011. Case COMP/M.794 - Coca-Cola/Amalgamated Beverages GB, Commission decision of 22 January 1997.
viii
table of merger decisions
ix
Case COMP/M.833 - Coca-Cola Company, The/Carlsberg A/S, Commission decision of 11 September 1997. Case COMP/M.3436 - Continental/Phoenix, Commission decision of 26 October 2004. Case COMP/M.6166 - Deutsche Börse/NYSE Euronext, Commission decision of 1 February 2012. Case COMP/M.3868 - DONG/Elsam/Energi E2, Commission decision of 14 March 2006. Case COMP/M.3030 - Eaton/Delta, Commission decision of 23 January 2003. Case COMP/M.5549 - EDF/Segebel, Commission decision of 12 November 2009. Case COMP/M.3440 - ENI/EDP/GDP, Commission decision of 9 December 2004. Case COMP/M.1225 - Enso/Stora, Commission decision of 25 November 1998. Case COMP/M.3696 - E.ON/MOL, Commission decision of 21 December 2005. Case COMP/M.1383 - Exxon/Mobil, Commission decision of 29 September 1999. Case COMP/M.2016 - France Telecom/Orange, Commission decision of 11 August 2000. Case COMP/M.5046 - Friesland Foods/Campina, Commission decision of 17 December 2008. Case COMP/M.4180 - Gaz de France/Suez, Commission decision of 14 November 2006. Case COMP/M.3304 - GE/Amersham, Commission decision of 21 January 2004. Case COMP/M.2220 - GE/Honeywell, Commission decision of 3 July 2001. Case COMP/M.3083 - GE/Instrumentarium, Commission decision of 2 September 2003. Case COMP/M.4561 - GE/Smiths/Aerospace, Commission decision of 23 April 2007. Case COMP/M.619 - Gencor/Lonrho, Commission decision of 24 April 1996. Case COMP/M.4215 - Glatfelter/Crompton Assets, Commission decision of 20 December 2006. Case COMP/M.4731- Google/DoubleClick, Commission decision of 11 March 2008. Case COMP/M.938 - Guinness/Grand Metropolitan, Commission decision of 15 October 1997. Case COMP/M.2503 - HBG/Ballast Nedam/Baggeren JV, Commission decision of 3 August 2001. Case COMP/M.2946 - IBM/PWC Consulting, Commission decision of 23 September 2002. Case COMP/M.3062 - IBM/Rational, Commission decision of 20 February 2003. Case COMP/M.4747 - IBM/Telelogic, Commission decision of 5 March 2008. Case COMP/M.4000 - Inco/Falconbridge, Commission decision of 4 July 2006. Case COMP/M.4094 - Ineos/BP Dormagen, Commission decision of 10 August 2006. Case COMP/M.4734 - Ineos/Kerling, Commission decision of 30 January 2008. Case COMP/M.5984 - Intel/McAfee, Commission decision of 26 January 2011. Case COMP/M.4874 - Itema/Barcovision, Commission decision of 4 August 2008. Case COMP/M.4381 - JCI/VB/FIAMM, Commission decision of 10 May 2007.
x
table of merger decisions
Case COMP/M.3687 - Johnson&Johnson/Guidant, Commission decision of 25 August 2005. Case COMP/M.6266 - Johnson&Johnson/Synthes, Commission decision of 18 April 2012. Case COMP/M.308 - Kali+Salz/MDK/Treuhand, Commission decision of 14 December 1993. Case COMP/M.623 - Kimberly-Clark/Scott, Commission decision of 16 January 1996. Case COMP/M.5141 - KLM/Martinair, Commission decision of 17 December 2008. Case COMP/M.4057 - Korsnäs/AD Cartonboard, Commission decision of 12 May 2006. Case COMP/M.5644 - Kraft Foods/Cadbury, Commission decision of 6 January 2010. Case COMP/M.4525 - Kronospan/Constantia, Commission decision of 19 September 2007. Case COMP/M.2978 - Lagardère/Natexis/VUP, Commission decision of 7 January 2004. Case COMP/M.5020 - Lesaffre/GBI UK, Commission decision of 11 July 2008. Case COMP/M.5440 - Lufthansa/Austrian Airlines, Commission decision of 28 August 2009. Case COMP/M.5335 - Lufthansa/SN Airholding, Commission decision of 22 June 2009. Case COMP/M.4187 - Metso/Aker Kvaerner, Commission decision of 12 December 2006. Case COMP/M.190 - Nestlé/Perrier, Commission decision of 22 July 1992. Case COMP/M.1355 - Newell/Rubbermaid, Commission decision of 13 January 1999. Case COMP/M.2876 - Newscorp/Telepiù, Commission decision of 2 April 2003. Case COMP/M.4942 - Nokia/Navteq, Commission decision of 2 July 2008. Case COMP/M.4781 - Norddeutsche Affinnerie/Cumerio, Commission decision of 23 January 2008. Case COMP/M.2499 - Norske Skog/Parenco/Walsum, Commission decision of 21 November 2001. Case COMP/M.5830 - Olympic/Aegean Airlines, Commission decision of 26 January 2011. Case COMP/M.3796 - Omya/Huber PCC, Commission decision of 19 July 2006. Case COMP/M.3216 - Oracle/Peoplesoft, Commission decision of 26 October 2004. Case COMP/M.5529 - Oracle/Sun Microsystems, Commission decision of 21 January 2010. Case COMP/M.2256 - Philips/Agilent Health Care Solutions, Commission decision of 2 March 2001. Case COMP/M.3191 - Phillip Morris/Papastratos, Commission decision of 2 October 2003. Case COMP/M.1882 - Pirelli/BICC, Commission decision of 19 July 2000. Case COMP/M.3732 - Procter&Gamble/Gillette, Commission decision of 15 July 2005. Case COMP/M.430 - Procter&Gamble/VP Schickedanz, Commission decision of 21 June 1994.
table of merger decisions
xi
Case COMP/M.1221 - Rewe/Meinl, Commission decision of 3 February 1999. Case COMP/M.4439 - Ryanair/Aer Lingus, Commission decision of 27 June 2007. Case COMP/M.5434 - Ryanair/Aer Lingus II, Commission decision of 23 January 2009. Case COMP/M.3943 - Saint-Gobain/BPB, Commission decision of 9 November 2005. Case COMP/M.5253 - Sanofi-Aventis/Zentiva, Commission decision of 4 February 2009. Case COMP/M.2283 - Schneider/Legrand, Commission decision of 10 October 2001. Case COMP/M.3848 - Sea-Invest/EMO-EKOM, Commission decision of 18 August 2006. Case COMP/M.6214 - Seagate/HDD Business of Samsung, Commission decision of 19 October 2011. Case COMP/M.2380 - SEB/FSB, Commission decision of 19 September 2001. Case COMP/M.290 - Sextant/BGT-VDO, Commission decision of 21 December 1992. Case COMP/M.4504 - SFR/Télé 2 France, Commission decision of 18 July 2007. Case COMP/M.2861 - Siemens/Drägerwerk/JV, Commission decision of 30 April 2003. Case COMP/M.3653 - Siemens/VA Tech, Commission decision of 13 July 2005. Case COMP/M.368 - SNECMA/TI, Commission decision of 17 January 1994. Case COMP/M.2690 - Solvay/Montedison-Ausimont, Commission decision of 9 April 2002. Case COMP/M.3431 - Sonoco/Ahlstrom, Commission decision of 6 October 2004. Case COMP/M.3333 - Sony/BMG, Commission decision of 19 July 2004. Case COMP/M.3333 - Sony/BMG, Commission decision of 3 October 2007. Case COMP/M.4919 - StatoilHydro/ConocoPhillips, Commission decision of 21 October 2008. Case COMP/M.4956 - STX/Aker Yards, Commission decision of 5 May 2008. Case COMP/M.5675 - Syngenta/Monsanto’s Sunflower Seed Business, Commission decision of 17 November 2010. Case COMP/M.4662 - Syniverse/BSG, Commission decision of 4 December 2007. Case COMP/M.2803 - Telia/Sonera, Commission decision of 10 July 2002. Case COMP/M.2416 - Tetra Laval/Sidel, Commission decision of 30 October 2001. Case COMP/M.4403 - Thales/Finmeccanica/Alcatel Alenia Space & Telespazio, Commission decision of 4 April 2007. Case COMP/M.4726 - Thomson Corporation/Reuters Group, Commission decision of 19 February 2008. Case COMP/M.3916 - T-Mobile Austria/Tele.ring, Commission decision of 26 April 2006. Case COMP/M.4854 - TomTom/Tele Atlas, Commission decision of 14 May 2008. Case COMP/M.5483 – Toshiba/Fujitsu HDD Business, Commission decision of 11 May 2009. Case COMP/M.4523 - Travelport/Worldspan, Commission decision of 21 August 2007. Case COMP/M.1736 - UIAG/Carlyle/Andritz, Commission decision of 17 November 1999.
xii
table of merger decisions
Case COMP/M.5658 - Unilever/Sara Lee Body Care, Commission decision of 17 November 2010. Case COMP/M.4404 - Universal/BMG Music Publishing, Commission decision of 22 May 2007. Case COMP/M.2498 - UPM-Kymmene/Haindl, Commission decision of 21 November 2001. Case COMP/M.6101 - UPM/Myllykoski and Rhein Papier, Commission decision of 13 July 2011. Case COMP/M.6228 - Vivendi/SFR, Commission decision of 9 June 2011. Case COMP/M.3512 - VNU/WPP/JV, Commission decision of 15 September 2004. Case COMP/M.1672 - Volvo/Scania, Commission decision of 15 March 2000. Case COMP/M.5907 - Votorantim/Fischer/JV, Commission decision of 4 May 2011. Case COMP/M.6203 - Western Digital Ireland/Viviti Technologies, Commission decision of 23 November 2011. Case COMP/M.4389 - WLR/BST, Commission decision of 5 December 2006.
ACKNOWLEDGMENTS
A number of people have assisted with the production of this book. Daniel Gore, Stephen Lewis and Andrea Lofaro are particularly indebted to their colleagues at RBB Economics for many of the materials drawn upon in its preparation. Special mention is deserved by Derek Ridyard, Simon Bishop and Simon Baker from whom we have learned a lot over the years – many of the insights into economic and policy issues contained in the book are due to them. Frances Dethmers is very grateful to her colleagues at Clifford Chance and, in particular Tony Reeves, Heleen Engelen and Adam Dawson for providing unconditional support. In addition, she would like to thank the ‘wrecking crew’ (Julianne O’Leary, Jürgen Schindler, Julia Holtz, Kylie Sturtz and Barbara Nijs) for making competition law that bit more exciting. Last but not least, thanks to Luc Peeperkorn (DG Comp) and Ali Nikpay (OFT) who were invaluable teachers at the beginning. We also thank Adrian Majumdar, Iestyn Williams, Sam Brown, Alan Crawford, Chris Doyle, Paul Hutchinson, Daria Prigioni and Toby Watt for critically reviewing various chapters and for providing invaluable insights, comments and suggestions. Joan de Solà-Morales assisted with the research and exposition of a number of merger cases, while the many charts and figures contained in the book would have looked far less professional without the help of Stéphanie Bouchet, Ina Esser and Alicia Fecci. Florentin Genthon, Julie Lassebie, Louise Lee and Michael Lewis all have our gratitude for spotting typos and suggesting innumerable drafting improvements. Both Monica Michiels van Kessenich and Chantal Roosseleers at Clifford Chance worked incredibly hard to improve the overall layout, including all references. Special thanks go to Nicholas Levy who agreed to write the foreword and to Simon Baxter, Fiona Carlin, Jeffrey Church, Claire Jeffs, Frédéric Louis and Frank Verboven for their encouraging feedback and kind endorsements. xiii
xiv
acknowledgments
We are also very grateful to Kim Hughes and her colleagues at Cambridge University Press for their enduring support throughout the project and for not losing patience despite countless missed deadlines over the last two years. Finally, we would like to thank Katy, Daria, Jane and Hilde for tolerating our absence during many evenings and weekends and for providing support throughout this project. We would like to dedicate this book to our parents.
FOREWORD
It is difficult now to recall the sense of expectation, uncertainty, and excitement that greeted the entry into force of the Merger Regulation in 1990. Among the many unanswered questions was how the Commission would define markets and analyse the substantive issues raised by reportable concentrations. Notwithstanding the explicit emphasis placed on competition-based criteria in the original text of the Merger Regulation, the use of economics and economists was in its infancy in the EU: few Commission officials had a background in economics; outside counsel were for the most part unfamiliar with economic theory and concepts; and economics was at the time applied only rarely in antitrust cases. Fundamental issues concerning the role of economic evidence in EU merger control were therefore unresolved and very real. Initially in respect of market definition and subsequently in connection with horizontal and non-horizontal effects, the Commission started to employ a more rigorous, quantitative, and economically orientated approach to the assessment of mergers and other forms of concentration, placing increasing reliance on firm evidence and solid investigative techniques that could be tested against what Joseph Schumpeter called ‘the cold metal of economic theory’. When, in 2003 and 2004, as part of the package of measures implemented in response to a series of reversals before the EU Courts, the Commission adopted the Horizontal Mergers Guidelines and appointed its first Chief Economist, the central role of economics and economists in the application of the Merger Regulation was confirmed, and decisions since then have been increasingly grounded in hard data and sound economics. This magnificent and comprehensive textbook introduces students, practitioners, economists, and regulators alike to the economic tools and methodologies used to assess concentrations under the Merger Regulation. The principal empirical techniques employed to define markets and to analyse unilateral effects, coordinated effects, vertical effects, and conglomerate effects are described with clarity and precision. Each xv
xvi
foreword
chapter explains – clearly, concisely, and thoughtfully – the underlying economic concepts before describing the Commission’s practice. Charts, tables, and graphs illustrate the applicable principles, and individual cases, including many of the leading decisions of the past decade – Nokia/Navteq, Sony/BMG, Google/DoubleClick, Statoil/JET, ABF/GBI Business, Oracle/Sun Microsystems and Gas Natural/Endesa – are given detailed analysis, drawing on the authors’ considerable experience and insights. And, most importantly perhaps, authoritative guidance is provided on those quantitative techniques that have been employed most successfully to address the wide array of substantive issues that have arisen in practice. No antitrust lawyer can aspire to practice EU merger control today without a sound knowledge of economics. This textbook provides that knowledge. It is a formidable achievement, a truly invaluable work. Daniel Gore, Stephen Lewis, Andrea Lofaro, and Frances Dethmers have authored a terrific antitrust reference book. Nicholas Levy December 2012
A COMMENT ON THE TEXT BY PROFESSOR JEFFREY CHURCH
This comprehensive and insightful volume is distinguished by its skillful and knowledgeable mix of relevant economic theory and practice of competition economics. The authors have been at the forefront of the economic revolution that has transformed merger enforcement policy in Europe in the last decade and this book reflects their inside knowledge and experience. In providing a comprehensive discussion of the economic concepts that underlie modern merger enforcement policy and in presenting and assessing the empirical techniques and methodologies used in practice to marshal and assess the evidence on competitive effects of a merger, whether horizontal, vertical, or conglomerate, the authors have provided a real and valuable service to the competition policy enforcement community worldwide. Their focus is clearly on the “what” of enforcement policy, with deft discussion of the actual use of the techniques in merger cases considered by the European Commission. This is a foundational text destined to be the “go to” reference for those new to the intricacies of the practice of merger policy enforcement, as well as experienced hands. Professor Jeffrey Church University of Calgary December 2012
xvii
1 Introduction
1.
Purpose and scope of this book
This book is intended to support and assist practitioners involved in the application of European competition law to mergers. Our principal intention is to provide lawyers who do not have a background in economics with an overview of the economic foundations of merger analysis, and of the analytical techniques and evidence used to appraise the competitive impact of mergers. We also hope that this book may be useful to economists who wish to gain an understanding of how economics is applied to merger assessment in practice. The goal is to assist readers to understand the economic concepts relevant to a particular case; to identify forms of economic analysis and evidence relevant to that case; to recognise what analyses and evidence would best be prepared by the merging parties and their advisers; and to evaluate critically economic evidence prepared on behalf of merging parties or by competition authorities. In line with this goal, the book is structured according to the types of issue that particular merger notifications may raise. Each issue, and the relevant forms of evidence and analyses, is discussed on a stand-alone basis. This approach is intended to allow the text to be used as a reference, with the reader able to consult the relevant section for the type of merger or question faced. Chapter 2 introduces the concept of market definition. Market definition is a conceptual framework for identifying the groups of firms, products and regions amongst which competitive interactions arise. As such, market definition is a central element of all antitrust investigations, including merger assessment, where market definition provides the starting point for investigating the impact of changes in firm ownership. Chapter 2 provides an overview of the evidence and analytical techniques used by the Commission to inform the assessment of market definition in merger investigations. 1
2
introduction
Chapter 3 discusses unilateral effects, a theory of harm that frequently arises in the context of horizontal mergers. Unilateral effects may lead to a lessening of competition where a merger brings together two firms whose products represent important substitutes for customers. Chapter 3 describes a range of considerations relevant to the assessment of unilateral effects, and surveys the various forms of evidence and analysis employed in practice to assess the scope for unilateral effects in horizontal mergers. Chapter 4 discusses coordinated effects, a less common theory of harm that may apply in the case of horizontal mergers. Coordinated effects arise where a merger changes market conditions such that firms may be better able to restrict competition between themselves via a tacit understanding of their joint interest in higher prices. Chapter 4 reviews the factors that economic theory predicts might make coordinated effects more or less likely, discusses the framework established by the Commission for the practical assessment of coordinated effects, and describes the forms of evidence that have been considered by the Commission within that framework. Finally, Chapter 5 considers non-horizontal mergers, which are distinguished from horizontal mergers by the fact that they do not concern products that customers would consider substitutes and therefore do not eliminate direct competitive constraints between firms. Non-horizontal mergers encompass vertical mergers between firms at different levels of a supply chain, and conglomerate mergers between firms active in different markets and different supply chains. While non-horizontal mergers will raise competition concerns less frequently than horizontal mergers, in some circumstances they may permit firms to engage in foreclosure, that is, behaviour that weakens rivals and consequently lessens competition. Chapter 5 discusses the various forms of foreclosure theory that may apply in non-horizontal mergers, the framework used by the Commission to assess these theories, and the types of evidence considered in such assessments. In preparation for this material, this introduction provides an overview of the framework within which mergers are assessed in Europe, focusing on the role of economic evidence and analysis. Section 2 starts by setting out the regulatory standard for merger assessment in the European Union and the procedure by which the European Commission investigates and rules upon notified transactions. Section 3 goes on to provide an introductory discussion of the role of economics in European merger analysis, focusing on the increased usage and importance of economic analysis over the last five years.
legal framework and commission procedure
3
Section 4 provides an introduction to the core economic principles that underpin merger control. It discusses the central concepts of industrial organisation, the academic field that provides the basis for economic analysis of firm behaviour and market structure. The practical application of industrial organisation theory to real-world firms and industries is the essence of the economic analysis of mergers. Theory alone is usually not sufficient to reach a firm conclusion on the likely impact of a merger on competition. While economic theory will often provide an indication of the direction of relationships between variables (for instance the relationship between a product’s price and demand for that product), it is generally not able to provide an indication of the strength of those relationships. The strength of relevant economic relationships must be assessed on an ad hoc basis for the industry and firms involved in each merger investigation using empirical evidence. This often requires the use of econometrics, a field concerned with the use of mathematics and statistics to connect economic theory with empirical data. This is the subject of Section 5, which provides a brief introduction to empirical economic evidence and its application to merger assessment.
2.
Legal framework and Commission procedure 2.1 The Merger Regulation
The legal basis for the regulatory supervision of corporate mergers and acquisitions in Europe is provided by the EC Merger Regulation (the ‘Merger Regulation’).1 The Merger Regulation applies to concentrations, defined as covering ‘operations bringing about a lasting change in the control of the undertakings concerned and therefore in the structure of the market’, a definition that encompasses joint ventures as well as mergers and acquisitions.2 The Merger Regulation acknowledges that such concentrations ‘are to be welcomed to the extent that they are in line with the requirements of dynamic competition and capable of increasing the competitiveness of European industry, improving the conditions of growth and raising the standard of living in the Community’.3 However, the Merger Regulation 1
2
Council Regulation (EC) No. 139/2004 of 20 January 2004 on the control of concentrations between undertakings (the EC Merger Regulation), OJ L24/1, 29 January 2004. 3 Merger Regulation, para. 20. Ibid., para. 4.
4
introduction
also notes that mergers should be permitted only in so far as they do not ‘result in lasting damage to competition’, and that the European Union asserts a legal basis for ‘governing those concentrations which may significantly impede effective competition in the common market or in a substantial part of it’.4 The Merger Regulation affords the European Commission exclusive jurisdiction over transactions that bring about ‘significant structural changes, the impact of which on the market goes beyond the national borders of any one Member State’.5 Transactions affecting individual Member States fall within the purview of the applicable national competition authorities.6 The Merger Regulation limits assessment by the Commission to mergers that meet specified turnover thresholds, concentrations that meet these thresholds being referred to as having a ‘Community dimension’.7 A concentration has a Community dimension if:8 (a) the combined aggregate worldwide turnover of all the undertakings concerned is more than EUR 5 000 million; and (b) the aggregate Community-wide turnover of each of at least two of the undertakings concerned is more than EUR 250 million, unless each of the undertakings concerned achieves more than two-thirds of its aggregate Community-wide turnover within one and the same Member State.
From an economic perspective, the most important element of the Merger Regulation is that concerning the competitive assessment of mergers. The Merger Regulation establishes the concept of a significant impediment to effective competition (‘SIEC’) as the criterion against 4 7 8
5 6 Merger Regulation, para. 5. Ibid., paras. 8 and 9. Ibid., para. 8. Ibid., paras. 9 and 10. Ibid., Art. 1, para. 2. Alternatively, a concentration has a Community dimension where:
(a) the combined aggregate worldwide turnover of all the undertakings concerned is more than EUR 2 500 million; (b) in each of at least three Member States, the combined aggregate turnover of all the undertakings concerned is more than EUR 100 million; (c) in each of at least three Member States included for the purpose of point (b), the aggregate turnover of each of at least two of the undertakings concerned is more than EUR 25 million; and (d) the aggregate Community-wide turnover of each of at least two of the undertakings concerned is more than EUR 100 million, unless each of the undertakings concerned achieves more than two-thirds of its aggregate Community-wide turnover within one and the same Member State. See ibid., Art. 1, para. 3.
legal framework and commission procedure
5
which concentrations are to be assessed. It provides that ‘any concentration which would significantly impede effective competition, in the common market or in a substantial part of it, should be declared incompatible with the common market’.9 The concept of an SIEC represented a departure from the standard set out in the previous regulation governing merger control in Europe (‘the 1989 Merger Regulation’).10 The 1989 Merger Regulation prohibited any concentration which ‘creates or strengthens a dominant position as a result of which effective competition in the common market or in a substantial part of it would be significantly impeded’.11 It should be apparent that the criterion for prohibition established in the Merger Regulation is a broadening of the equivalent condition in the 1989 Merger Regulation. Both refer to a significant or substantial impediment to effective competition, but the 1989 Regulation includes an additional requirement for prohibition that a dominant position be created or strengthened that is absent from the 2004 Merger Regulation. The dominance provision in the 1989 Regulation was held by some commentators to give rise to a ‘gap’, whereby the Commission would be legally prevented from prohibiting mergers with the potential to harm consumers through a lessening of competition in oligopolistic industries in which no individual firm was dominant.12 While the Commission had, via its case law, created the concept of collective dominance with which it prohibited mergers under the 1989 Merger Regulation in industries not characterised by a single dominant firm, this concept depended on establishing scope for firms to coordinate their actions via collusive behaviour.13 The enforcement gap was held to arise in the case of mergers taking place in markets in which there was no realistic prospect of coordinated
9 11 12
13
10 Merger Regulation, para. 25. Council Regulation (EEC) No. 4064/89. 1989 Merger Regulation, para. 24. An oligopolistic market is one characterised by a small number of competing suppliers, such that those firms are able to influence price and take into account the behaviour of rivals when determining their own strategies (as distinct from the textbook model of pure competition, in which individual firms have no such influence but must charge the price determined by the market or make zero sales). In practice, almost all markets of interest to competition law are oligopolies. For an articulation of the view that the Commission’s decision to prohibit the Airtours/ First Choice merger on collective dominance grounds possibly represented an attempt to sidestep a lacuna in the 1989 Merger Regulation see M. Motta (1999) ‘EC Merger Policy, and the Airtours case’, available at http://people.exeter.ac.uk/maf206/motta_1999.pdf.
6
introduction
behaviour, and no single dominant firm, but nonetheless the prospect of a lessening of competition via unilateral effects. The 2004 Merger Regulation was introduced to address this possible gap in enforcement, and comments on the concept in the following terms.14 In view of the consequences that concentrations in oligopolistic market structures may have, it is all the more necessary to maintain effective competition in such markets. Many oligopolistic markets exhibit a healthy degree of competition. However, under certain circumstances, concentrations involving the elimination of important competitive constraints that the merging parties had exerted upon each other, as well as a reduction of competitive pressure on the remaining competitors, may, even in the absence of a likelihood of coordination between the members of the oligopoly, result in a significant impediment to effective competition … The notion of ‘significant impediment to effective competition’ … should be interpreted as extending, beyond the concept of dominance, only to the anti-competitive effects of a concentration resulting from the non-coordinated behaviour of undertakings which would not have a dominant position on the market concerned.
However, since the introduction of the new substantive test, the Commission’s enforcement practice suggests that the alleged ‘gap’ in merger control under the dominance test was smaller than initially believed.15 An article published in April 2006 by Lars-Hendrik Röller, former Chief Economist, and Miguel de la Mano, Deputy Chief Economist, found no horizontal merger in the sample they reviewed that was a clear cut ‘gap case’, although it was suggested that one vertical merger (E.ON/MOL)16 might constitute a gap case that may not have been challenged under the 1989 Merger Regulation.17 Since then, gap features have emerged in only a very small number of cases. In particular, in T-Mobile Austria/Tele.ring,18 which brought together the second and fourth largest network operators on the
14 15
16 17
18
Merger Regulation, para. 25. See A. Lofaro and D. Ridyard, ‘The Role of Economics in European Merger Control’, published in N. Levy, European Merger Control Review – A Guide to the Merger Regulation (LexisNexis, 2011). Case COMP/M.3696 – E.ON/MOL, Commission decision of 21 December 2005. See L.-H. Röller and M. de la Mano, ‘The Impact of the New Substantive Test in European Merger Control’ (2006) 2(1) European Competition Journal 9. Case COMP/M.3916 – T-Mobile Austria/Tele.ring, Commission decision of 26 April 2006 (2007 OJ L88/44).
legal framework and commission procedure
7
Austrian mobile telecommunications sector, the Commission concluded that anti-competitive effects would occur despite the fact that the merged entity would account for only one-third of the market. Similarly, in EDF/ Segebel,19 which involved EDF’s acquisition of a majority stake in SPE, the second largest electricity operator in Belgium, the Commission required the parties to make significant divestments despite a combined share in the Belgian electricity wholesale market of only 10–20%. In both cases, the Commission considered that remedies were warranted since the proposed transactions as originally notified would have eliminated rivals whose competitive importance was significantly understated by their market shares.
2.2 Procedure for notification and assessment It is beyond the scope of this book to give a complete description of merger notification and assessment under the Merger Regulation. In this section we instead seek to provide an overview of the various elements of the Commission procedure and timelines for those unfamiliar with the process. Mergers falling under the Merger Regulation are assessed by the Directorate-General for Competition (‘DG COMP’). The Merger Regulation establishes a system of mandatory notification for concentrations with a Community dimension, stating that:20 undertakings should be obliged to give prior notification of concentrations with a Community dimension following the conclusion of the agreement, the announcement of the public bid or the acquisition of a controlling interest. Notification should also be possible where the undertakings concerned satisfy the Commission of their intention to enter into an agreement for a proposed concentration and demonstrate to the Commission that their plan for that proposed concentration is sufficiently concrete, for example on the basis of an agreement in principle, a memorandum of understanding, or a letter of intent signed by all undertakings concerned, or, in the case of a public bid, where they have publicly announced an intention to make such a bid, provided that the intended agreement or bid would result in a concentration with a Community dimension.
19 20
Case COMP/M.5549 – EDF/Segebel, Commission decision of 12 November 2009. Merger Regulation, para. 34.
8
introduction
Mergers are normally notified to the Commission via a standard notification document template known as Form CO.21 The purpose of the Form CO is to provide the Commission with the information concerning the transaction that it requires to conduct its competitive assessment. In particular, the Form CO calls for information about the parties and their ownership, details of the concentration, relevant internal documents, the definition of the relevant markets affected by the transaction, and the structure of and competitive conditions within those markets. A merger notification becomes effective, setting the administrative timetable underway, on the date on which the Commission receives a Form CO that it accepts as complete. In principle, therefore, the Form CO may represent the first point of contact between the parties and the Commission. In practice, however, parties involved in transactions that may raise competition issues will often engage in pre-notification discussions with the Commission. During pre-notification the Commission will provide feedback on drafts of the Form CO, which allows the parties to identify areas in which further argumentation, evidence or analysis would assist in dispelling competition concerns. Once a notification has become effective, the Commission commences a 25-working-day Phase I initial examination. This Phase I process may be extended to 35 days if the parties offer commitments (also known as remedies) aimed at addressing potential competition concerns. During this investigation the Commission must form a view on whether the transaction raises serious doubts as to its effect on competition. In order to reach such a view, the Commission will take account of information from the parties contained in the Form CO and responses to supplementary questionnaires issued during the course of the Phase I process. Importantly, the Commission will augment this material with its own market investigation, which will canvass the views of and collect evidence from third parties (particularly customers, but also rivals) active in the markets affected by the transaction. Given the limited timescale, the Commission will typically not pursue detailed or sophisticated economic analyses during a Phase I investigation but instead focus on identifying and weighing the merits of potential concerns flagged by the information received from the parties and market participants. It will, nonetheless, engage with and consider detailed economic evidence and analyses put forward by the parties, 21
The Form CO is provided as an annex to Commission Regulation (EC) No. 802/2004 implementing Council Regulation (EC) No. 139/2004.
legal framework and commission procedure
9
particularly where such analyses have been discussed with the Commission during pre-notification. It is therefore feasible for parties to bring complex and sophisticated economic evidence into play during Phase I merger proceedings. The Phase I investigation period ends in a decision under Article 6 of the Merger Regulation. In the significant majority of cases, the Commission concludes that the transaction does not raise serious doubts, leading to an Article 6(1)(b) approval decision. Where such a finding depends on commitments offered by the parties, the Article 6(1)(b) decision will be issued in conjunction with Article 6(2), making the approval conditional on compliance with those commitments. An approval decision will close the merger assessment and permit the parties to complete the transaction. Alternatively, if the Commission has serious doubts as to the impact of the transaction at the end of its Phase I investigation, it will issue an Article (6)(1)(c) decision, triggering a Phase II investigation. The basic timetable for the Phase II review is 90 working days, although this is extended to 105 working days where the parties offer commitments later than 55 days into the Phase II process, and/or may be extended by 20 working days with the agreement of the parties. The Phase II process normally involves a detailed and evidenceintensive review of specific competition concerns identified during the Phase I process. The Phase II investigation provides sufficient time for the Commission to engage fully with detailed and sophisticated economic analyses and evidence provided by the parties, and for it to undertake its own economic analyses based on information both from the parties and its market investigation. Where third party information plays a part in the Commission’s investigation there may be scope for the use of a data room process, whereby the parties’ advisers, under suitable confidentiality conditions, are permitted to review and comment on data and analyses employed by the Commission’s staff and/or provided by third parties. The Phase II process ultimately leads to a decision as to the transaction’s compatibility with competition law under Article 8 of the Merger Regulation. Typically, the Commission will reach one of three decisions at the end of a Phase II investigation: unconditional clearance under Article 8(1); clearance subject to commitments addressing identified competition concerns under Article 8(2); or, in the case of a transaction deemed incompatible with the common market, prohibition under Article 8(3).
10
introduction
Phase I: Initial Examination
Concentration does not raise serious doubts as to compatibility with common market
Article 6(1)b: Approval or
25 working day deadline (35 if commitments offered) commences on receipt of complete notification.
Article 6(2): Approval subject to commitments
Concentration raises serious doubts as to compatibility with common market
Article 6(1)c: Reference
Concentration compatible with common market Phase II: Initiation of Proceedings 90 working day deadline (105 if commitments offered later than 55 days from initiation of proceedings), plus 20 working day extension if agreed by parties, commences on date of Article 6(1)(c) decision
Article 8(1): Approval
Concentration compatible with common market subject to commitments Article 8(2): Approval subject to commitments
Concentration incompatible with common market
Article 8(3): Prohibition
Figure 1.1
European merger review process under the Merger Regulation
Where the Commission intends to prohibit a transaction or clear it only subject to commitments, such a decision will be preceded by a statement of objections (‘SO’), which sets out the Commission’s basis for such a decision. The SO provides the parties with an opportunity to respond to the Commission’s arguments prior to the decision, both in writing and at an oral hearing. Figure 1.1 summarises the European merger review process under the Merger Regulation.
legal framework and commission procedure
11
Table 1.1 Conclusion of merger cases in which the Commission has reached a decision under Article 6 (from 2004 to 2011)
Article 6 decisions, of which … … cleared unconditionally … cleared subject to commitments … Phase II proceedings initiated, of which… … withdrawn … cleared unconditionally … cleared subject to commitments … prohibited
Number of cases
Proportion of Article 6 decisions (%)
2,453 2,271 109 73 15 27 28 3
92.6 4.4 3.0 0.6 1.1 1.1 0.1
Source: European Commission statistics.
Table 1.1 summarises the conclusion of merger cases in which the Commission has reached a decision under Article 6 over the eight years from 2004 to 2011. These data show that the vast majority of mergers investigated by the Commission are cleared at Phase I without commitments. Only around 7.5% of mergers that reach the end of the Phase I process require commitments or a more detailed investigation under Phase II. It is also notable that a sizeable proportion, around one-third, of cases that proceed to Phase II are ultimately cleared without conditions, demonstrating that serious doubts identified in the Phase I process may nevertheless be overcome by a more detailed review. Finally, the table shows that only a tiny fraction of the transactions notified to the Commission are ultimately prohibited on competition grounds.22
22
In addition to the three prohibitions observed over the eight years shown in the table, it may be inferred that some fraction of the fifteen notifications withdrawn in Phase II may otherwise have resulted in a prohibition decision; similarly, some proportion of the fifty notifications withdrawn during Phase I (not shown in the table) may have ultimately led to prohibition. Nonetheless, the vast majority of transactions considered by the Commission are ultimately cleared, and most of these without commitments.
12
introduction
3.
The role of economics in EU merger control
Merger analysis in the EU has undergone a radical transformation in the last decade with an increased reliance on economics in all areas of EU competition law. This approach is often referred to as ‘effects-based’, whereby the focus is on the impact that particular market structures and conduct have on outcomes for consumers. This is contradistinct to a ‘form-based’ approach that regulates competition according to prescribed rules defining acceptable and unacceptable forms of firm behaviour and market structure. An effects-based competition enforcement regime necessarily comes at a cost in terms of the legal certainty afforded by a form-based approach, whereby firms can immediately understand whether a particular act (e.g. offering loyalty rebates, merging with a supplier) will or will not risk intervention on competition grounds; the associated need for case specific investigation and analysis also places a greater burden on both firms and regulators. Nonetheless, it is now generally acknowledged that the application of insights provided by economics to a particular case is more likely to avoid both under- and over-intervention by competition regulators, and so produce outcomes that best serve consumers. In the context of merger appraisal, this change in enforcement policy has most obviously been embodied in the shift away from market shares as a tool for the assessment of unilateral effects in horizontal mergers in favour of a greater focus on the economic concepts of substitutability and closeness of competition. Arguably, however, the move towards effectsbased analysis has had its most significant impact in the assessment of non-horizontal and coordinated effect theories of harm. In both of these areas there has been a departure from the use of checklists and per se prohibitions over the past ten years. From a practitioners’ point of view, the move towards effects-based enforcement has led to increased reliance on economic analysis and evidence in European merger control, both on the part of the Commission and merging parties. DG COMP created the post of Chief Competition Economist (CCE) in 2003 as part of a reform programme intended to address evidential and analytical failings highlighted by the General Court’s annulments of the Airtours/First Choice,23 Tetra Laval/Sidel24 and 23 24
Case COMP/M.1524 – Airtours/First Choice, Commission decision of 22 September 1999. Case COMP/M.2416 – Tetra Laval/Sidel, Commission decision of 30 October 2001.
basic economic concepts
13
Schneider/Legrand25 prohibition decisions. The CCE heads a group of specialist competition economists, the Chief Economist’s Team (CET), that provides the Commission with in-house expertise in the application of economic theory and econometric techniques to antitrust enforcement. In merger control the CET supports investigation case teams via two primary functions. First, the CET provides independent internal review of proposed clearance, reference and prohibition decisions (the ‘checks and balances’ function). Second, the CET participates in and assists case teams with ongoing merger assessments (the ‘support’ function).26 Via the first of these roles the CET has been instrumental in improving the quality of Commission decision making in recent years, by subjecting case teams to internal scrutiny in terms of both the economic logic of their arguments and the strength of the supporting evidence. On a day-to-day basis, however, it is the second function that most directly affects merger practitioners. The support role encompasses the allocation of a CET member to join the case team on mergers that raise complex economic or evidential issues. Such allocations may be made at the request of either the case team or the CCE. As a result, CET economists typically play an important role in the most complex and contentious cases, particularly those under or likely to enter Phase II review.
4.
Basic economic concepts
Throughout this book we will introduce the particular economic concepts and theories that are relevant to each type of merger and theory of harm at the beginning of the applicable chapters. There are, however, a number of basic economic concepts that underpin all merger assessment. In this section we provide an overview of these concepts, focusing on the demand curve, the nature and role of firms’ costs, rational profit-maximising behaviour, and the concept of market power. Competition economics lies within the broader field of industrial organisation (‘IO’), which is the branch of microeconomics that deals
25 26
Case COMP/M.2283 – Schneider/Legrand, Commission decision of 10 October 2001. See L.-H. Röller and P. A. Buigues, ‘The Office of the Chief Competition Economist at the European Commission’ (Commission, May 2005).
14
introduction
with the behaviour of firms and the structure of markets. The fundamental topic of interest in industrial organisation is the analysis of how rational firms go about maximising their profits given the behaviour of consumers and other firms. This question is addressed through consideration of firms’ cost conditions, demand conditions and the nature of interaction between firms. Each firm is assumed to maximise its profits by reference to its own cost conditions and the demand it faces, taking account of other firms’ profitmaximisation decisions. Competition economics is built upon the analysis of individual firms’ incentives and behaviour. In particular, merger analysis derives from the analysis of how firms’ current incentives and behaviour, a result of current cost and demand conditions, might be expected to change following a consolidation of two or more separate firms under common control. As will be demonstrated below, the most significant change in incentives following a merger arises through changes in the demand curve facing firms involved with or affected by the merger. The following sections provide a brief overview of the basic economic concepts relevant to competition analysis in general, and to merger analysis in particular.
4.1 The demand curve The demand facing a firm derives from a wide range of factors. These include the price of the firm’s goods, the characteristics of the product, the price and attributes of other suppliers’ products, consumers’ income and consumers’ tastes and preferences. Economists summarise these factors in the concept of the demand curve. A demand curve is a description of the volume of sales that a firm will make for each price level it sets, holding other factors (such as product characteristics and competitors’ prices) constant. Figure 1.2 plots an illustrative demand curve for WidgetCo, a firm producing widgets. The x-axis shows the quantity (q) of widgets sold by WidgetCo, while the y-axis shows the price (p) charged by WidgetCo. The demand curve reflects the relationship between these two variables, illustrating the number of units that WidgetCo would sell at each price level; or, analogously, the price that would have to be set to sell a given number of widgets. The demand curve is downward sloping, reflecting the fact that higher prices lead to lower volumes and vice versa. This negative relationship
basic economic concepts
15
p( )
4 3 2 D
200
Figure 1.2
300
400
q
Illustrative demand curve
holds true for (practically) all demand curves.27 The demand curve for WidgetCo is a ‘linear’ demand curve (i.e. a straight line), with every €1 change in price leading to a 100 unit change in sales. We shall see in the discussion of profit maximisation below that firms combine this information on the relationship between price and sales volume with information on their own costs to decide on the level of price (and therefore output) to offer. Firms’ behaviour thus depends on the characteristics of the demand curve faced. 27
In theory, demand may be upward sloping for some goods, such that higher prices increase demand for the good. One (theoretical) class of such goods is known as ‘Giffen goods’. These are a special case of so called ‘inferior goods’ (goods for which demand increases as income decreases). If an inferior good represents a sufficiently large part of a consumer’s total expenditure, an increase in the price of such a good may reduce income available for expenditure to such an extent that it causes an increase in demand for the good itself (for which demand is inversely related to income). Giffen goods are of theoretical interest because, despite their apparent absurdity, they cannot be ruled out by the axioms of rational choice that underpin microeconomic theory. However, they are of little or no practical relevance. Another class of products that may appear to have upward sloping demand curves are known as ‘Veblen goods’. These are goods that attract higher demand at higher prices because consumers associate higher prices with higher perceived quality. It has been conjectured that luxury goods such as perfume may represent Veblen goods. It could, however, be argued that, while a bottle of perfume sold at €100 may attract more demand than a chemically identical perfume sold at €10, the two are not the same product due to the extent to which the price is an element of the product’s characteristics for such items. It is not clear, for instance, that a bottle of perfume normally retailing at €100 would lose customers if its price were to be reduced by 5% as a promotional offer.
16
introduction
A critical element of the demand curve for competition analysis is the elasticity of demand, normally represented by the Greek symbol ε. The own-price elasticity of demand measures the responsiveness of demand for a good to changes in its price, that is, how many sales the firm can expect to win (lose) as a result of a given reduction (increase) in price.28 The own-price elasticity of demand is defined as the ratio of the percentage change in demand to the percentage change in price causing that change in demand. This can be expressed via the following formula, where Δq is the change in the quantity demanded, q is the quantity originally demanded, Δp is the change in price and p is the original price: ε ¼ ðΔq=qÞ=ðΔp=pÞ Thus, to give an example, if a 10% increase in price would lead to a 5% reduction in sales, the elasticity would be 5%/10%, or 0.5. Note that, because demand is downward sloping, the own-price elasticity of demand will always be a negative number. By convention, however, elasticity is often expressed in absolute terms (i.e. the minus sign is omitted), such that the own-price elasticity of demand in this case would be described as 0.5. It is important to note that elasticity will normally be different at different points along the demand curve.29 Indeed, because elasticity is calculated on the basis of proportionate, rather than absolute, changes in price and volume, own-price elasticity will always differ at different points on a linear demand curve. This can be seen with the illustrative linear demand curve shown for WidgetCo above. At a price of €2, WidgetCo would be selling 400 units. A €1 price increase to €3 would lead to a 100 unit reduction in sales, to 300. The elasticity at this point is therefore 25%/50%, which is 0.5. At a price of €3, WidgetCo would be able to sell 300 units. A €1 price increase at this price level would again lead to a 100 unit reduction in demand. However, the proportionate price increase is now 33%, while the proportionate volume change is 33%. The elasticity at this point is thus 1. 28
29
In addition to own-price elasticity, the cross-price elasticity of demand, which measures the responsiveness of one good’s demand to changes in another good’s price, is also relevant to competition analysis. See the discussion of demand estimation in the context of market definition at Chapter 2. The exception is the constant elasticity, or isoelastic, demand curve, which is defined according to a specific shape such that elasticity does not change as one moves along the demand curve. The constant elasticity demand curve is discussed further at Appendix B.
basic economic concepts
17
Finally, at a price of €4, WidgetCo would sell 200 units. Given that the demand curve is a straight line, a €1 price increase will again lead to a 100 unit demand fall (from 200 to 100 units). The percentage increase in price is now 25%, whereas the percentage decrease in demand is now 50%, implying that the elasticity at this point is 2. Thus, while the relationship between price and sales for WidgetCo is linear, each €1 change in price being associated with a 100 unit change in demand, the elasticity changes according to the starting point at which it is measured. WidgetCo will therefore face different demand elasticity depending upon the price that it is charging. Specifically, at prices lower than €3, it can be seen that WidgetCo has an elasticity of demand that is less than one in absolute terms, whereas at prices higher than €3, it has an elasticity of demand that is greater than one in absolute terms. At a price of €3 exactly, its elasticity of demand is equal to one in absolute terms. Where elasticity is less than one in absolute value (i.e. less negative than 1), demand is referred to as being ‘inelastic’, or ‘relatively inelastic’.30 Demand will be relatively inelastic where a given percentage change in price will lead to a less than proportionate percentage change in volume. Where elasticity is greater than one in absolute value (i.e. more negative than 1), demand is called ‘elastic’ or ‘relatively elastic’.31 In regions of relatively elastic demand, changes in price will lead to more than proportionate changes in volume. At the point between these two regions, demand elasticity of (minus) one is referred to as ‘unitary elasticity’. It is useful at this stage to introduce the concept of marginal revenue. Marginal revenue is the rate of change of revenue with respect to sales volumes. A decision by a firm to increase output by one unit will affect its revenues in two ways. The direct effect will be a boost to its revenue as it will earn extra revenue on the incremental unit it sells. However, as an indirect effect, it will suffer a loss in revenues as it will have to lower the price of all units it sells in order to make that incremental sale. Whether marginal revenue is positive or negative will depend on which of these two effects is larger. 30
31
Perfectly inelastic demand (ε ¼ 0) refers to a vertical demand curve, at which the same output is demanded irrespective of price. Perfectly elastic demand (ε ¼ ∞) arises where the demand curve is horizontal, which implies that there exists a price at which the firm can sell any number of units, but that any price above this level will result in zero sales.
18
introduction
Recall that we noted that where demand elasticity is less than one in absolute terms, a given percentage change in price will lead to a smaller percentage change in volume. Therefore, a price reduction in the region where demand is inelastic will reduce total revenue, as the lower price will more than offset the increase in sales. We can therefore see that where demand is inelastic, the relationship between output and revenue is negative, meaning that marginal revenue is negative. Conversely, where demand elasticity is greater than one in absolute terms, a given percentage change in price will lead to a larger percentage change in volume. Therefore, a price reduction in the region where demand is elastic will increase total revenue, as the increase in sales will more than outweigh the reduction in price. Thus, where demand is elastic, the relationship between output and revenue is positive, meaning that marginal revenue is positive. Marginal revenue therefore falls as output increases. More specifically, the marginal revenue curve will lie above zero where demand is elastic, cross zero at the point where demand elasticity is unitary, and lie below zero where demand is inelastic. Figure 1.3 illustrates the relationship between demand elasticity, marginal revenue and total revenue for a linear demand curve. The chart shows that total revenue is maximised at the point of unitary elasticity, or where marginal revenue is equal to zero.32 This does not imply, however, that firms will aim to operate at the point on their demand curve at which elasticity is equal to one. Profit-maximising firms must also take into account the costs of production associated with each price and output level. Even without considering costs, however, it can be noted that a firm will always operate in the elastic portion of its demand curve, or, equivalently, where marginal revenue is positive. This can be seen by considering the choices available to a firm operating in the inelastic portion of its demand curve. Where demand is inelastic, revenue is unambiguously increasing with price because a given percentage increase in price will bring about a smaller percentage decrease in volume. Given that costs would not be expected to increase with reductions in output, it follows that increasing total revenue while reducing volume must increase profitability. As such, whenever a firm faces inelastic demand (or where marginal revenue is negative) it has an incentive to increase price, such 32
The characteristic that revenue is maximised when marginal revenue is zero is not only the case for linear demand curves but holds more generally.
basic economic concepts
19
p
Elastic demand
Unitary elasticity
Inelastic demand
MR>0
MR50% (Wholesale gas)5 90–100% (Gas storage)6 70–75% (Wholesale and retail electricity7)
ENI/EDP/GDP8
>50% (Wholesale gas9)
Gaz de France/Suez10
80–100% (Various markets for gas11)
SFR/Télé 2 France12
>50%13
Input foreclosure concerns The transaction was considered to be likely to confer on DONG the ability and incentive to raise rival’s costs by increasing gas storage tariffs in Denmark. The Commission considered that the new entity would have the ability and incentive to foreclose access to wholesale gas to its competitors in the gas retail market, as well as to discriminate against its competitors in granting access to gas storage. Foreclosure concerns were also considered to arise in the wholesale and retail electricity markets. The Commission concluded that the merged entity would likely have an incentive to foreclose its downstream competitors by raising the price (or lowering the quality) of their gas supplies. The merged entity was deemed to have greater opportunities and incentives to increase its downstream competitors’ costs for ancillary services and balancing power. The Commission concluded that the merged entity would have both the ability and incentive to disadvantage downstream pay TV rivals through a range of discriminatory measures notwithstanding the minimal additional share of Télé 2 for pay TV distribution (50%’ in those instances where the EC did not provide a specific estimate but its explanations (i.e. ‘clearly dominant’, ‘quasi-monopoly’, etc.) or figures were suggesting that the share in question was well above 50%. The market share of the main rival is provided when available and relevant. 2 Case COMP/M.3868 – DONG/Elsam/Energi E2, Commission decision of 14 March 2006. 3 ‘DONG’s overall dominance in the Danish gas sector is partly due to its storage monopoly in Denmark’: see ibid., para. 298. 4 Case COMP/M.3696 – E.ON/MOL, Commission decision of 21 December 2005. 5 ‘MOL WMT is currently dominant on all the markets for the wholesale supply of gas in Hungary (gas traders and large power plants) and its dominant position is not likely to be threatened in the short to medium term’: see ibid., para. 284. 6 ‘MOL Storage is the only company owning gas storage facilities in Hungary and therefore the only one able to offer gas storage services’: see ibid., para. 479. 1
7
See ibid., paras. 629–30. Case COMP/M.3440 – ENI/EDP/GDP, Commission decision of 9 December 2004. 9 ‘By allowing EDP to acquire the supplier of the main input used for the production of electricity, and hence to integrate upstream with the only supplier of natural gas in Portugal, the operation is likely to change immediately and in the near future, the conditions of competition on the wholesale market and to result in the strengthening of EDPs dominant position.’ See ibid., para. 366. 10 Case COMP/M.4180 – Gaz de France/Suez, Commission decision of 14 November 2006. 11 See ibid., paras. 134, 139, 144, 145, 148, 157, 159 and 160. 12 Case COMP/M.4504 – SFR/Télé 2 France, Commission decision of 18 July 2007. 13 ‘Vivendi’s very substantial position on the upstream and intermediate markets would enable it to boost the proprietary package of SFR/Télé 2 considerably by giving it attractive or differentiated content (channels or broadcasting rights) which is not accessible to other DSL operators or which is available to them only on terms less advantageous than those given to SFR/Télé 2’: see ibid., para. 91. 14 Case COMP/M.4726 – Thomson Corporation/Reuters Group, Commission decision of 19 February 2008. 15 The Commission assessed shares in the four segments – Fundamentals, Estimates, Broker Research or Economic Time Series – separately but dedicated a section of the decision to the combined foreclosure concerns of the four abovementioned services. It noted that ‘[60–70]% of the competitors specifically redistribute Fundamentals, Estimates, Broker Research or Economic Time Series from Thomson or Reuters. It should be noted that the largest competitors, but Bloomberg, all redistribute one or several of these four specific content sets from the notifying parties.’ See ibid., para. 423. 16 Case COMP/M.5675 – Syngenta/Monsanto’s Sunflower Seed Business, Commission decision of 17 November 2010. 17 See ibid., para. 164. 8
vertical mergers: input foreclosure
403
increase following the merger, the prospect of further downstream integration and therefore of the loss of a sizeable, if not total, sensor sales for Uster will limit its incentive to exploit it by raising prices.[sic]’36 In Nokia/Navteq,37 discussed in more detail in Section 3.1.3 below, the Commission considered whether competition from Tele Atlas (and Garmin) would undermine any attempt by Navteq to raise downstream rivals’ costs by increasing upstream market power.38 3.1.2.2 The importance of the input As discussed above, anticompetitive input foreclosure will be feasible only if the input in question is important for the production of the relevant product(s) downstream, either because it represents a substantial proportion of costs or is in some sense essential to the downstream activity. If the price of the input represents only a small part of total variable costs, even a substantial price increase is unlikely to have a significant impact on the rival’s ability to compete. In Nokia/Navteq,39 the Commission found that the importance of the input (navigable digital maps) varied between the downstream markets to which it was supplied (see Section 3.1.3 for further discussion). In Thomson Corporation/Reuters Group,40 the Commission measured the dependency of competitors on Thomson/Reuters upstream content via the share of downstream competitors’ desktop workstations that carried the parties’ content. In some segments, these shares were substantial, contributing to the Commission’s conclusion that foreclosure strategies would be viable post-merger. 3.1.2.3 Potential foreclosure strategies In most cases, the focus of the Commission has been on pricing strategies, whereby the merged entity charges higher prices to downstream rivals. Nonetheless, there is in principle a wide array of strategies beyond simple refusal to supply and increases in upstream prices that vertically integrated firms may potentially employ to undermine downstream competition. Such strategies include the following. 36 37 38
39 40
Ibid., para. 69. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. A similar issue was considered in TomTom/Tele Atlas. See Case COMP/M.4854 – TomTom/Tele Atlas, Commission decision of 14 May 2008. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. Case COMP/M.4726 – Thomson Corporation/Reuters Group, Commission decision of 19 February 2008.
404
non-horizontal mergers
41 • Sales limitation: in Arsenal/DSP, the Commission argued that the merged entity could increase prices and restrict the amount of benzoic acid supplied to rivals downstream, forcing downstream rivals to purchase their remaining requirements from distributors or non-EEA producers at higher prices. 42 • Quality degradation: the Commission held in Nokia/Navteq that the merged entity would have the ability not only to increase prices but also to degrade the quality of maps supplied to downstream rivals, delay rivals’ access to updated maps, or reserve innovative features for Nokia.43 Similarly, in SFR/Télé 2 France44 the Commission’s market investigation suggested that the merged entity may be able to weaken the competitive pressure exerted by rivals in the downstream market for the distribution of pay TV by offering exclusive or preferential broadcasting rights to its own downstream channels.45 In Nokia/ Navteq the Commission also argued that the merged entity may engage in other tactics, such as delaying responses to requests for price quotations or important technical information.46 This approach demonstrates that quality degradation need not apply to the product itself but may occur in respect of related products or services such as aftersales services, add-ons, or optional features.
3.1.2.4 Other factors that may affect the ability to foreclose In addition to considerations relating to the upstream firm’s market power and the importance of the input, there may be other case-specific factors affecting an integrated firm’s ability to pursue foreclosure strategies postmerger. The Commission has accepted in several cases that long-term contractual obligations may prevent a merged firm from raising prices or modifying other terms and conditions at least within the next few years following the merger. Whilst such inability may not be permanent, it may provide rivals with sufficient time to find or develop alternative sources 41 42 43 44 45 46
Case COMP/M.5153 – Arsenal/DSP, Commission decision of 9 January 2009, para. 283. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. Ibid., para. 264. Case COMP/M.4504 – SFR/Télé 2 France, Commission decision of 18 July 2007. Ibid., para. 92. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008; see also Case COMP/M.4403 – Thales/Finmeccanica/Alcatel Alenia Space & Telespazio, Commission decision of 4 April 2007, para. 256, where the Commission discussed several delaying tactics.
vertical mergers: input foreclosure
405
of supply. The Commission has noted the use of long-term contracts as a potential impediment to input foreclosure on several occasions. For instance, in TomTom/Tele Atlas,47 the Commission noted: ‘Tele Atlas’s ability to foreclose its downstream competitors is limited by the long term contract that Garmin has concluded with NAVTEQ, which protects Garmin against price increases and guarantees yearly price decreases at least until 2015.’48 Nevertheless, the existence of such contracts is typically not regarded by the Commission as sufficient in and of itself to allay foreclosure concerns.49 Conversely, the existence of short-term contracts or contracts that are close to renewal may be regarded as a factor potentially increasing an integrated firm’s ability to foreclose. In Thomson Corporation/Reuters Group50 the Commission held that: ‘contracts generally cover a short period … and most renegotiations are already under way … Therefore, the notifying parties will have the ability to foreclose competitors in this market.’51 A vertically integrated firm may be unable to reduce supply where it is already economically or legally committed to a certain level of capacity utilisation, or where the products in question are already in development. In Thales/Finmeccanica/Alcatel Alenia Space & Telespazio,52 the Commission’s analysis of input foreclosure focused on the pre-award stage of bidding competitions for satellite and satellite subsystems since ‘[p]ost award input foreclosure strategies (essentially delays in the implementation of TWT supply agreements) appear less likely, as they would be more easily detectable (in particular, customers can require a contractual right to send resident experts to follow the production process at their supplier) and would trigger pre-defined penalty payments.’53
47 48 49
50
51 52
53
Case COMP/M.4854 – TomTom/Tele Atlas, Commission decision of 14 May 2008. Ibid., para. 208. See, for instance, Case COMP/M.5153 – Arsenal/DSP, Commission decision of 9 January 2009, paras. 286–8, or Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008, para. 329. Case COMP/M.4726 – Thomson Corporation/Reuters Group, Commission decision of 19 February 2008. Ibid., para. 372. Case COMP/M.4403 – Thales/Finmeccanica/Alcatel Alenia Space & Telespazio, Commission decision of 4 April 2007. Ibid., para. 255.
406
non-horizontal mergers
Apart from these contractual obligations towards customers, a merged entity’s ability to engage in foreclosure strategies may be undermined by its own suppliers or its JV partners. In Thomson Corporation/Reuters Group,54 the parties submitted that broker companies that provide the necessary contents for some of the merged entity’s services would not agree to any foreclosure strategy that would reduce their output or exposure.55 The Commission rejected this argument, however, as it was unclear whether broker companies had the means and incentives to oppose a foreclosure strategy. Finally, regulatory intervention can also play an important role and affect the ability of the merged entity to engage in input foreclosure. In E.ON/MOL,56 the Commission considered the pricing conditions imposed by energy regulators to be an important factor determining the ability of the merged entity to engage in input foreclosure practices. The Commission took into consideration expected changes in the regulatory framework (regulated prices were expected to be suppressed in July 2007), and concluded that these would ease the implementation of a foreclosure strategy by the merged entity.57
3.1.3 Case study: Nokia/Navteq – assessment of ability to engage in input foreclosure During the Nokia/Navteq58 merger investigation, the Commission undertook a detailed assessment of whether the merged entity might have the ability to engage in input foreclosure. Navteq supplied digital maps that are used in Personal Navigation Devices (PNDs), particularly mobile phone handsets that offer navigation services. The Commission considered whether the merged entity would be able to foreclose firms active in a number of downstream markets either by increasing the price and/or reducing the quality of maps supplied to rival mobile phone handset manufacturers.59 The Commission considered three conditions that are necessary for the merged entity to have the ability to engage in foreclosure. 54
55 56 57 58 59
Case COMP/M.4726 – Thomson Corporation/Reuters Group, Commission decision of 19 February 2008. Ibid., paras. 224–5. Case COMP/M.3696 – E.ON/MOL, Commission decision of 21 December 2005. Ibid., para. 363. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. Ibid., para. 270.
vertical mergers: input foreclosure
407
First, the Commission considered whether the merged entity would have market power in the upstream market for navigable digital maps. In reaching its view that the merged entity did indeed have market power, the Commission highlighted structural features of the market for navigable digital maps. Specifically, it noted that Navteq had a market share of around 50% and faced only one competitor, Tele Atlas, whose navigable digital map databases had a similar coverage and quality level.60 The Commission also noted the high gross margins earned by Navteq. Whilst accepting that high gross margins do not necessarily indicate prices in excess of the competitive level in an industry with high fixed costs and low marginal costs, the Commission stated that Navteq’s ‘ability to price well above marginal cost is indicative of market power.’61 Furthermore, it pointed out that non-navigable digital map databases would be only imperfectly substitutable to navigable digital map databases for downstream competitors, and these services would be imperfect substitutes to navigation services based on navigable maps.62 Second, the Commission considered the importance of the input for downstream competitors, noting that, irrespective of its share of downstream cost, an input may be important on account of being a critical component. The Commission found that the importance of the input varied across the downstream markets to which navigable digital maps were provided and considered each in turn. Regarding the market for mobile handsets, the Commission found that digital maps accounted for a relatively small share (0–15%) of the total costs of producing mobile handsets with navigation functionality, but that, where navigation services were provided on a particular handset, digital maps constituted a critical component.63 Given that navigation services are only one feature among others that determine a customer’s choice of handset (such as the handset’s music or video player), the Commission regarded it as unclear whether navigable digital map databases represented a critical input in the market for mobile handsets. As noted in our example above, if an input represents only 10% of downstream firms’ production costs then even a 50% increase in the price of that input would only increase downstream costs by 5%. Regarding the market for navigation applications on mobile handsets, the Commission found that digital maps accounted for a significant share of the costs of navigation software providers (whether these were selling 60 61
Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008, para. 273. 62 63 Ibid., para. 275. Ibid., para. 295. Ibid., para. 285.
408
non-horizontal mergers
to mobile network operators or commercialising them directly) and that digital maps were a critical input.64 Third, the Commission considered whether timely and effective counterstrategies would be available to downstream rivals.65 The first such counterstrategy that the Commission considered was the scope for entry in the upstream market for provision of navigable digital map databases. The Commission considered that entry was unlikely to occur and dismissed this as a potential constraint on the merged entity’s ability to foreclose.66 The Commission also considered the role of intermediaries that hold map database and navigation software licences from Tele Atlas or Navteq to constrain the merged entity’s ability to foreclose.67 Garmin, a PND manufacturer and software developer, had concluded a long-term contract with Navteq (until 2015 with the option for Garmin to extend until 2019). Under the terms of its contract, Garmin could embed Navteq maps in its own PNDs or integrate them into navigation software that it could sell to any type of user, including for navigation services on mobile handsets.68 The Commission concluded that Garmin could represent, to a certain extent, a credible supplier of navigable digital map databases for handset manufacturers or mobile network operators (at least in the short term).69 In terms of the example illustrated in Section 3.1.1 above, Garmin could be considered to play a similar role to Firm C. Correspondingly, if Garmin’s offer of navigation software were sufficiently attractive to rival mobile handset manufacturers, this could undermine the ability of Tele Atlas to exert market power even if Navteq had engaged in a strategy of total foreclosure. Furthermore, the Commission assessed the degree of buyer power enjoyed by mobile network operators, and found that whilst they exerted a degree of power on Nokia, it was unclear whether they would be able to counter any strategy of Navteq aimed at foreclosing Nokia’s downstream competitors.70 Ultimately, the Commission’s findings regarding the incentive to foreclose (discussed in Section 3.2.3) enabled it to leave open its assessment
64 65
66 67
Ibid., para. 290. In practice, it will be apparent that the factors considered by the Commission under the heading of counterstrategies could perhaps be considered within the assessment of Navteq’s upstream market power. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008, para. 300. 68 69 70 Ibid., para. 306. Ibid., para. 306. Ibid., para. 315. Ibid., para. 325.
vertical mergers: input foreclosure
409
of whether the merged entity would have the ability to foreclose. In particular, the Commission concluded that it was unclear whether the merged entity would have the ability to engage in input foreclosure given the uncertainty over whether navigable digital map databases are a critical input in the market for handsets and the potential for timely and effective counterstrategies (particularly relating to the role that Garmin might play).71 Also left open in the Commission’s assessment is an important question regarding whether the merged entity could credibly commit to a strategy of foreclosure.72 The parties argued that Navteq could not commit to foreclose Nokia’s downstream competitors since Navteq would have an incentive to undercut Tele Atlas slightly to make a profitable sale without impacting its downstream sales. Since Tele Atlas would be aware of this possibility, and it could do nothing to prevent it from occurring, it would refrain from raising its price in the first place.73 This commitment problem is examined in the next subsection as part of our discussion of the assessment of the incentive to engage in input foreclosure.
3.2 Incentive to engage in input foreclosure The preceding subsection established the factors that will determine whether a vertically integrated firm will have the ability to implement a strategy of anti-competitive foreclosure against unintegrated downstream rivals. In this subsection we consider under what conditions, even where that ability exists, an integrated firm will have an incentive to pursue such a strategy. A vertical merger alters the incentives of the upstream division because it will now take account of the impact of its pricing decisions on its downstream division. This may give rise in some circumstances to an incentive to engage in input foreclosure to benefit the merged entity’s downstream division even if this comes at some cost to the upstream division. In short, the incentive to engage in input foreclosure depends on the effect of the hypothesised foreclosure strategy on the profits of the merged entity as a whole. An assessment of a firm’s incentive to foreclose therefore requires a separate evaluation of the negative effect on profits of the upstream 71 72
Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008, para. 329. 73 Ibid., para. 271. Ibid., para. 299.
410
non-horizontal mergers
division of the firm and the positive effect on profits of the downstream division resulting from the foreclosure strategy.74 The negative effect on upstream profits results from the fact that input foreclosure involves the sacrifice of sales to downstream rivals.75 The positive effect on downstream profits arises indirectly as a result of the impact of foreclosure on the price charged by downstream rivals. To the extent that downstream rivals increase their prices, this may either cause some sales to divert to the downstream division of the merged firm or allow it to raise the price it charges final consumers (or some combination of the two). A number of interrelated factors determine whether the positive effect on downstream profits will outweigh the negative effect on upstream profits leading to an incentive to foreclose. First, this balance depends on the relative sizes of the margin earned on forgone upstream sales and the margin earned on incremental downstream sales. If the margin on upstream sales is relatively high, this will imply that the opportunity cost of forgoing upstream sales is relatively high. A larger increase in downstream sales will therefore be required to compensate for the lost upstream sales in order for the foreclosure strategy to be incrementally profitable. Conversely, if the margin on downstream sales is relatively high, this will imply that fewer incremental downstream sales are required to justify any sacrifice of upstream sales. In the example set out in Section 3.2.1 below we consider further how relative margins earned on upstream and downstream sales affect the critical increase in demand for the downstream division required to make a foreclosure strategy incrementally profitable. Second, for any given ratio of upstream margin to downstream margin, the incentive to engage in input foreclosure will depend on the increase in demand it delivers to the downstream division. This will depend on the effect of the foreclosure strategy on the costs of
74
75
See EC Non-horizontal Merger Guidelines, para. 40: ‘Essentially, the merged entity faces a trade-off between the profit lost in the upstream market due to a reduction of input sales to (actual or potential) rivals and the profit gain, in the short or longer term, from expanding sales downstream or, as the case may be, being able to raise prices to consumers.’ In this case of total foreclosure all sales to unintegrated rivals are lost. In the case of partial foreclosure, fewer sales are made to unintegrated rivals at a higher price. However, assuming that the upstream firm was profit-maximising pre-merger, the total profits of the upstream division (viewed in isolation) will necessarily be lower if it increases its price as part of a partial foreclosure strategy.
vertical mergers: input foreclosure
411
downstream rivals (which in turn depends on the factors affecting the ability to foreclose, discussed in Section 3.1 above, such as the importance of the input); the extent to which any cost increase is passed on to consumers by downstream rivals in the form of higher prices; and the extent to which those downstream consumers react by substituting to the products of the merged entity’s downstream division (i.e. the extent to which the merged entity’s products are substitutable for those of the firms that are the subject of the foreclosure strategy).76 In the example below we consider further how these considerations are taken into account in assessing the actual increase in downstream demand that a merged entity could expect to result from input foreclosure. In principle, both the critical increase in downstream demand required to make a foreclosure strategy profitable and the actual increase in demand resulting from that strategy will also depend on (i) any increase in downstream margins due to the relaxation of the constraint imposed by downstream competitors or (ii) any increase in downstream margins due to efficiencies. We consider how these factors affect the analysis of the incentive to foreclose in the example below. Finally, an important issue that will be considered below is whether the merged entity can credibly commit to a strategy of foreclosure.
3.2.1 Incentive to engage in input foreclosure: an example In Section 3.1.1 we considered a stylised example in which a merger between an upstream component producer (Firm A) and a downstream gadget producer (Firm 1) gave rise to an ability on the part of the merged entity (Firm A1) to engage in total foreclosure. In what follows, we use the same illustrative example to examine the incentive the merged entity may have to engage in such a strategy. We assume again that three Firms, A, B and C, produce an input that is a component to a gadget produced by two downstream Firms, 1 and 2, and that the component represents 100% of the gadget’s marginal cost of production.77 As before, the three upstream suppliers incur no fixed costs but differ with respect to the marginal cost they incur in the production of the component as set out in Table 5.1 in Section 3.1.1 above. 76
77
The profitability of this strategy is also contingent on the downstream division having the necessary spare capacity to meet any increase in demand the foreclosure strategy generates. We refer to ‘Firm A’ and ‘Firm 1’ when they are unintegrated. Post-merger these entities are referred to as ‘Division A’ and ‘Division 1’ of the merged entity ‘Firm A1’.
412
non-horizontal mergers A
P = 20
C
P = 20
1
Figure 5.5
B
2
Illustration of pre-merger scenario
We now assume that the gadgets produced by downstream firms are differentiated, so that even if Firm 2’s price is higher than Firm 1’s price, some consumers may prefer Firm 2’s gadget. As in our previous example, pre-merger Firm A is able to win the business to supply both Firm 1 and Firm 2 by offering a price of 20. (See Figure 5.5.)78 Suppose that, at this price, Firm A sells 100 units to Firm 1 and 100 units to Firm 2. Further assume that Firm 1 and Firm 2 each sell 100 gadgets at a unit price of 40. It is now possible to obtain the pre-merger profits of both Firm A and Firm 1. Firm A sells a total of 200 units and earns a margin of 10 on each unit (the price of 20 less its marginal cost of 10) for a total profit of 2,000. Firm 1 sells a total of 100 units and earns a margin of 20 on each unit (the price of 40 less its marginal cost of 20) for a total profit of 2,000. We now consider the post-merger profits of both Division A and Division 1 on the assumption that the merged entity continues to supply Firm 2 at a price of 20 but now supplies Division 1 at a price of 10 (equal to its marginal cost). This situation is illustrated in Figure 5.6. We further assume that both Division 1 and Firm 2 maintain their price of gadgets at 40.79 Division A now sells 100 units to Firm 2 at a 78
79
Its price is constrained by the presence of B, which would be willing to supply at a price at least as high as 20. As explained above, we assume that Firm A wins the business at a price of 20 even though it has to offer a price an arbitrarily small amount below 20 in order to make its offer strictly preferred to Firm B’s offer. It is important to note that, in line with the discussion of the pro-competitive effect of vertical mergers at Section 2.1 above, the effect of the reduction in Division 1’s marginal cost will be to provide an incentive to reduce the price of gadgets and sell a greater volume but we abstract away from this effect at this stage. The effect of taking this into account is discussed below.
vertical mergers: input foreclosure A
P = 10
C
P = 20
1
Figure 5.6
B
413
2
Illustration of post-merger scenario without input foreclosure
margin of 10 per unit and 100 units to its downstream Division 1 at zero margin for a total profit of 1,000. Division 1 now sells 100 units and earns a margin of 30 on each unit (the price of 40 less its marginal cost of 10) for a total profit of 3,000. The combined profit of the merged entity Firm A1 is therefore equal to 4,000.80 Having established the profit that the merged entity can earn without foreclosure, we can now consider the merged entity’s profit if it were to engage in foreclosure. To the extent that the profits for the merged entity from engaging in foreclosure are greater than 4,000, the merged entity would have an incentive to foreclose. First note that the profits of Division A as a result of engaging in total foreclosure drop to zero. It makes no sales to Firm 2 and makes sales to Division 1 at marginal cost. The profits earned by Division 1 are less straightforward to calculate. As noted in Section 3.1.1, the strategy of total foreclosure will increase Firm 2’s cost from 20 to 30, the price at which it is supplied by Firm B.81 The increase in Firm 2’s marginal cost will cause it to increase the price it charges consumers for gadgets by some amount. Assuming (for the time being) that division 1’s price for gadgets remains at 40, the increase in Firm 2’s price will cause some of its consumers to purchase from Division 1 instead. This situation is illustrated in Figure 5.7.
80
81
Note that this is equal to the combined pre-merger profit of Firm A and Firm 1; the reduction in the component price to 10 is offset by resultant wider margins in the sale of gadgets, leaving overall profitability, absent foreclosure, unchanged. Firm B is now able to win the bidding contest to supply Firm 2 by offering a price of 30. Its price is constrained by the presence of Firm C, which would be willing to supply at a price at least as high as 30.
414
non-horizontal mergers A
P = 10
C
P = 30
1
Figure 5.7
B
2
Illustration of post-merger scenario with input foreclosure
We now consider the critical increase in Division 1’s demand required to ensure that the profits of Firm A1 are at least as high as 4,000, the merged entity’s profit supposing that it does not engage in foreclosure. Recall that the margin earned on each sale by Division 1 is 30. In order to generate profits of 4,000, its sales will need to be at least as high as 134 units (since if its sales were only 133 units, its profits would be equal to 3,990 or 133 * 30). The critical increase in Division 1’s sales is therefore equal to 34 units.82 In short, the loss of upstream profits of 1,000 requires the downstream firm to increase sales by 34 units given its margin of 30 (since 34 * 30 ¼ 1,020). It can be seen that the critical increase in Division 1’s sales depends on the ratio of the margin earned by Division A on forgone component sales to the margin earned by Division 1 on incremental gadget sales. Division A would have earned a margin of 10 on each forgone component sale to Firm 2 but earns a margin of 30 on each incremental gadget sale made by Division 1. Therefore the overall effect on the merged entity’s profits is positive if, for each incremental gadget sale, the upstream division has had to forgo no more than 3 component sales. Given that in this example total foreclosure implies the loss of 100 component sales, the strategy is profitable so long as more than 33 (≈ 100 component sales * 10 upstream margin/30 downstream margin) incremental gadget sales are made. Had the margin on the forgone component sales been higher, the critical increase in gadget sales required to make the foreclosure strategy profitable would also have been higher.
82
If it were possible to produce a fraction of a gadget, it could generate profits of exactly 4,000 by selling 33 and one-third of a gadget.
vertical mergers: input foreclosure
415
Whether the total foreclosure strategy is incrementally profitable depends on whether the actual increase in Division 1’s sales is greater than this critical value. This will depend on: • the extent to which the increase in Firm 2’s marginal cost of 20 to 30 is passed on to end-consumers of gadgets; and • the extent to which consumers of gadgets faced with an increase in Firm 2’s price switch to gadgets supplied by Firm 1 (i.e. the degree of substitutability between Firm 1’s and Firm 2’s gadgets). The higher the proportion of the component cost increase that is passed on to consumers in the form of higher gadget prices, the lower the substitutability between the downstream firms’ gadgets required for the actual increase in Division 1’s demand to exceed the critical amount. Conversely, the higher the substitutability between the downstream products, the lower the proportion of the cost increase that is passed on by Firm 2 required for the actual demand increase to exceed the critical demand increase.83 In Section 3.1.1 we noted that the extent to which Firm 2’s cost is increased by the total foreclosure strategy would depend on the increase in the component price faced by Firm 2 and the importance of the component in the overall marginal cost of producing gadgets. Specifically, the greater the competitive constraint imposed by Firm C on Firm B’s price after Division A’s refusal to supply, the lower will be the price increase brought about by the foreclosure strategy. The lower the proportion of the marginal cost represented by the component, the lower will be the effect on marginal cost of any given price increase. If it were the case that foreclosure would only increase the marginal cost of Firm 2 from 20 to 25 (rather than to 30) then this would increase either the substitutability between the downstream products or the proportion of the cost increase passed on (or some combination of the two) required in order for the actual increase in Division 1’s demand to exceed the critical increase. 3.2.1.1 Downstream price increases So far we have assumed that Division 1’s gadget price remains constant at 40 post-merger. The benefit from foreclosure therefore derives purely from an increase in the volume 83
It should be noted that these factors are themselves interrelated since the extent to which an increase in Firm 2’s costs will lead it to increase price will partly depend on the extent to which any price increase will cause it to lose sales to Firm A1.
416
non-horizontal mergers
of gadgets sold by Division 1. However, any increase in the price of Firm 2’s gadgets brought about by the foreclosure strategy may in fact give rise to an incentive for Firm 1 to increase its price due to the relaxation of the competitive constraint imposed by Firm 2. To the extent that Division 1 is able to increase its margin on gadget sales, this will reduce the critical increase in downstream demand required to make the foreclosure strategy profitable. For example, suppose that due to the relaxation of the constraint imposed by Firm 2 the merged entity finds it profitable to set a price of 50 for gadgets and hence earns a margin per sale of 40. Given that Division A would have earned a margin of 10 on each forgone component sale, the loss in profits for the upstream division is 1,000. However, even if Division 1 gains no incremental sales, the increase in downstream profits resulting from the higher margin is 1,000. This is because the margin is increased by 10 (from 30 to 40) across 100 unit sales resulting in an increase in profit of 1,000, which just offsets the profits lost upstream as a result of the foreclosure strategy. This means that if the foreclosure strategy increases the price from 40 to 50, the critical increase in demand required to make the foreclosure strategy profitable is zero. Importantly, while an increase in Division 1’s gadget price will reduce the critical increase in demand required to render foreclosure profitable, any such increase will also reduce the actual increase in demand observed compared to a situation in which it held its price constant. This is because some consumers of Firm 2’s gadget that might otherwise have switched following Firm 2’s price increase may in the face of higher prices for the merged entity’s gadgets remain with Firm 2 rather than switching. It should be noted that the result that the critical increase in demand is zero in this example, arises because we have supposed that the merged entity’s downstream price increases substantially (by 25%). This demonstrates the principle that the critical increase in downstream demand required to make foreclosure profitable depends on downstream price increases. However, as noted below, the assessment of the critical increase in demand will often necessarily be carried out on the basis of pre-merger margins, in which case the critical increase is necessarily greater than zero. 3.2.1.2 Efficiencies So far this example has not considered the role of efficiencies. Recall that pre-merger, Firm 1 faced a component cost of 20 and sold gadgets for 40. The merger reduces the (internal) component
vertical mergers: input foreclosure
417
cost faced by Division 1 to 10 (Division A’s marginal cost of supply). Division 1 will use this input cost reduction in part to reduce its downstream gadget price and in part to increase its margins. This will have two effects, each of which may serve to increase the incentive to foreclose: first, the increase in downstream margins will reduce the critical increase in demand required for foreclosure to be profitable (fewer gadget sales are now required to compensate for a given loss of component sales); and, second, Division 1’s lower price will raise the actual increase in demand observed via switching from Firm 2. 3.2.1.3 Use of pre-merger margin information Attempting to take account of how an input foreclosure strategy and efficiencies may interact to affect the post-merger downstream margin greatly complicates any assessment of both the critical and actual increase in the downstream unit’s demand. As was the case in Nokia/Navteq,84 which is discussed further at Section 3.2.3 below, it is typically necessary to rely on pre-merger margin information to assess the critical increase in the downstream unit’s demand and to assume the downstream unit will hold prices constant when assessing the actual increase in demand it would enjoy. Whilst an assumption that margins remain the same in the downstream market post-merger greatly simplifies the assessment of both the critical and actual increase in demand, and therefore the assessment of the overall incentive to foreclose, this assumption comes with a cost. This approach abstracts from the question that is ultimately of interest for the competitive assessment: what effect will foreclosure have on the downstream price for gadgets that consumers ultimately pay? Nonetheless, by making this assumption, it is often possible to gain significant insights into the profitability of a hypothesised foreclosure strategy for the purpose of assessing a firm’s incentive to engage in foreclosure. Indeed, consideration of pre-merger margins in the upstream and downstream markets along with estimates of the degree of switching between the downstream firms that would be induced by a change in relative prices may be sufficient to rule out the incentive to foreclose in many cases.85 For example, if downstream margins are very low compared to upstream 84 85
Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. See Chapter 3 for a discussion of the techniques for assessing substitutability, which include the use of survey analysis, customer switching analysis and econometric estimation of demand.
418
non-horizontal mergers
margins, this may give significant comfort that it is not worth sacrificing valuable upstream sales in the hope of gaining incremental downstream sales given a reasonable upper bound estimate of the substitutability of the downstream products. In that case, it might be concluded that foreclosure would not be profitable even if it were successful in raising downstream prices to some extent. As noted above, the relative size of upstream and downstream margins can provide insights into the critical increase in downstream demand for the merged firm that is required to make input foreclosure profitable. This needs to be compared to some measure of the actual increase in demand that would result from input foreclosure in order to determine whether foreclosure would be profitable. It is generally not possible to conclude on the profitability of input foreclosure solely by comparing upstream and downstream margins. Indeed, if downstream firms produce goods that are undifferentiated, this may imply that downstream margins are relatively low (due to intense competition) and therefore that the critical increase in downstream demand is relatively high. However, with limited differentiation amongst downstream products consumers are likely to be more sensitive to small increases in rivals’ prices brought about by the foreclosure strategy, implying that the actual increase in demand from a given foreclosure strategy would also be relatively high. This implies that there is no clear-cut relationship between the relative sizes of upstream and downstream margins and the incentive of a merged firm to engage in input foreclosure.86 3.2.1.4 The commitment problem The stylised example set out above has considered the incentives of a vertically integrated firm to raise rivals’ costs via a strategy of total foreclosure. However, we have not yet considered an important question about the incentive of a firm to engage in total foreclosure. Suppose that we assume as before that downstream margins will be unchanged by the merger (i.e. will remain at 30 per unit), giving rise to a critical increase in demand for Division 1 of 34 units. Suppose further that survey evidence of a high degree of substitutability between Firm 1’s and Firm 2’s gadgets leads to the 86
See R. Inderst and T. Valletti, ‘Incentives for Input Foreclosure’ (2011) 55(6) European Economic Review 820–31. This paper considers the incentives for a vertically integrated firm to foreclose downstream rivals and notes that pre-merger margins on the up- and downstream markets are not necessarily good predictors of the incentives for subsequent (i.e. post-merger) foreclosure.
vertical mergers: input foreclosure
419
conclusion that the actual increase in demand resulting from total foreclosure is likely to exceed 34 units (e.g. most respondents state that even a small increase in the price of Firm 2’s gadgets would cause them to switch to Division 1). We would then conclude that total foreclosure leads to higher profits than the benchmark case in which there is no foreclosure. However, can the merged entity credibly commit not to supply Firm 2? Consider that in this example A’s refusal to supply implies that Firm 2 must accept Firm B’s price of 30 (the price that just keeps Firm C from being able to profitably undercut Firm B). However, if Firm B offers a price of 30 to Firm 2, Firm A has an incentive to undercut this price by offering a price of 29. To see why this is the case, note that undercutting Firm B by offering to supply Firm 2 at a price of 29 only marginally undermines the total foreclosure strategy. In effect, relative to a situation in which it faces total foreclosure, Firm 2’s input price has fallen from 30 to 29. This will only have a ‘small’ impact on the price Firm 2 charges downstream and thus only a ‘small’ impact on the effectiveness of the merged entity’s foreclosure strategy in terms of generating incremental downstream profits for Division 1. However, undercutting Firm B in this way generates a ‘large’ increase in profits for Division A, equal to the margin it earns per component sale (19 ¼ 29–10) multiplied by the number of units that Firm 2 is willing to buy at the price of 29.87 On this basis, it follows that the total foreclosure strategy does not constitute a Nash Equilibrium (see Chapter 3, Section 2 for an explanation of Nash Equilibrium) as there exists a profitable deviation from the strategy of refusing to supply Firm 2 in the form of marginally undercutting Firm B’s price. However, Division A securing the business of Firm 2 at a price of 29 is also not an equilibrium. This is because Firm B will have an incentive to undercut Division A by offering Firm 2 a price of 28 (since this would result in it making sales at a price above its marginal cost). This process of undercutting will continue until Firm A is forced to supply Firm 2 at a price of 20, equal to Firm B’s marginal cost, in order to prevent Firm B from supplying Firm 2. This is the pre-merger price set by Firm A. Consequently, this logic implies that, unless Firm A can find some way of credibly committing not to supply Firm 2, it will not be able to engage in foreclosure.88 87
88
Note that this quantity will be lower than 100, the quantity that it purchased when the price was 20. The seminal paper by Ordover, Saloner and Salop that investigated the incentive for a merged entity to foreclose in a theoretical setting similar to our example was criticised in
420
non-horizontal mergers
One response to this argument is that a forward-looking merged entity will realise that entering into the price competition of the kind discussed above will lower the unintegrated downstream firm’s input price, which, in turn, will lead to lower downstream prices.89 However, this justification involves a conjecture that considerations that may apply in a dynamic framework are equivalent to the commitment assumption in the static framework.90 One way in which Division A may be able to commit not to supply Firm 2 is to take some action that makes its component incompatible with the gadget produced by Firm 2.91 In order to be effective, this strategy requires that the compatibility decision taken by the merged entity is not easily reversible (i.e. the cost of re-establishing compatibility between Division A’s components and Firm 2’s gadgets is reasonably large). The economic literature highlights that the ability of the merged entity to be able to commit not to supply its unintegrated downstream rivals, or to supply only at a certain price, is an important consideration for understanding how input foreclosure can arise.
3.2.2 Incentive to engage in input foreclosure: EU case law The Commission has conducted a detailed examination of the merged entity’s incentives to engage in input foreclosure strategies in a number of cases. This examination is usually based on economic evidence pertaining to the relevant factors that determine the trade-off between the profits lost in the upstream market(s) and the potential profits gained
89
90
91
a number of subsequent papers on the grounds that the model’s results are driven by the assumption that the merged entity would be able to commit not to supply. See J. Ordover, G. Saloner and S. C. Salop, ‘Equilibrium Vertical Foreclosure’ (1990) 80 American Economic Review 127–42. For a critique regarding commitment, see D. Reiffen, ‘Equilibrium Vertical Foreclosure, Comment’ (1992) 82 American Economic Review 694–7 and O. Hart and J. Tirole, ‘Vertical Integration and Market Foreclosure’, Brookings Papers on Economic Activity: Microeconomics (1990), 205–86. For discussion, see J. Church, ‘The Impact of Vertical and Conglomerate Mergers on Competition’ (Report for the Directorate General for Competition, Directorate B Merger Task Force, European Commission, 2004). See J. Ordover, G. Saloner and S. C. Salop, ‘Equilibrium Vertical Foreclosure, “Reply”’ (1992) 82 American Economic Review 693–704. See D. Reiffen and M. Vita, ‘Comment: Is There New Thinking on Vertical Mergers?’ (1995) 63 Antitrust Law Journal 917–41. See J. Church and N. Gandal, ‘Systems Competition, Vertical Merger and Foreclosure’ (2000) 9 Journal of Economics Management and Strategy 25–51.
vertical mergers: input foreclosure
421
in the downstream market(s). In Nokia/Navteq92 (see Section 3.2.3, below) and TomTom/Tele Atlas,93 the Commission sought to quantify the overall effect in a more comprehensive manner using econometric analysis to assess the actual diversion of demand downstream that could be expected to occur as a result of foreclosure. The Commission typically assesses the average gross margins (i.e. prices minus variable costs) of the relevant products upstream and downstream. If gross margins are considerably higher for upstream products than for downstream products, any given loss of sales upstream will need to be compensated by a greater increase in volumes downstream. For example, in Itema/Barcovision,94 the Commission noted, ‘The percentage margins are relatively small downstream and much higher upstream, which, all other things being equal, makes any extra revenues in the downstream winders market less likely to compensate for upstream losses than in a situation where downstream margins are high and upstream margins are small.’ 95 Similarly, in Thales/Finmeccanica/Alcatel Alenia Space & Telespazio,96 the Commission considered the merging parties’ margins at each level of the supply chain. The Commission found that margins for satellites (downstream) were significantly lower than margins for satellite components (upstream). The main source of this difference was that the downstream market was particularly competitive. This contributed to the conclusion that the incentives of the merged entity to forgo units sold upstream in order to win units sold downstream would be very limited.97 As noted above, given information on pre-merger margins it is in principle possible to calculate the critical increase in downstream sales needed to recoup the profit loss upstream, under the assumption that pre-merger margins are a good proxy for post-merger margins. Such a calculation was performed in Arsenal/DSP,98 where the Commission noted:
92 93 94 95 96
97 98
Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. Case COMP/M.4854 – TomTom/Tele Atlas, Commission decision of 14 May 2008. Case COMP/M.4874 – Itema/Barcovision, Commission decision of 4 August 2008. Ibid., para. 74. Case COMP/M.4403 – Thales/Finmeccanica/Alcatel Alenia Space & Telespazio, Commission decision of 4 April 2007. Ibid., para. 267. Case COMP/M.5153 – Arsenal/DSP, Commission decision of 9 January 2009.
422
non-horizontal mergers As the gross margin obtained by Velsicol/DSP in selling one tonne of dibenzoate plasticizer is approximately more than the gross margin it obtains when selling the quantity of benzoic acid necessary to make one tonne of di-benzoate plasticizers, the foreclosure strategy would be profitable if Velsicol/DSP could capture more than [50–60]% of the sales lost by a di-benzoate competitor. In other words, the new entity could afford to lose all sales of benzoic acid to a particular downstream competitor producing di-benzoate plasticizers, if it captures more than [50–60]% of the sales of di-benzoate plasticizers of this competitor. On the contrary, if Velsicol/DSP cannot capture more than [50–60]% of the sales of this competitor, it is unlikely that the foreclosure strategy would be profitable.99
Intuitively, the merged entity will find it more costly to engage in input foreclosure where it is unable to target the foreclosure strategy at competing products downstream. In Arsenal/DSP, one rival (Caffaro) purchased benzoic acid to produce both ketones and di-benzoate plasticisers and used more than 50% of this benzoic acid for the production of ketones. The merged entity was only active downstream in the production of di-benzoate plasticisers but could not discriminate between these two usages. The Commission therefore concluded on this basis, and also considering that Caffaro did not impose a strong constraint on Arsenal pre-merger, that the new entity would have little incentive to raise prices or restrict sales of di-benzoate plasticisers to this particular rival.100 The Commission has also considered the importance of each market’s contribution to the merged entity’s total revenues. For instance, in DONG/Elsam/Energi E2,101 one of the main reasons contributing to the Commission’s finding of an incentive to foreclose was the weight of each of the two activities (upstream and downstream) within total company sales. In particular, the Commission highlighted that ‘while the revenues from storage contribute only 5% of the DONG group’s EBITDA, its gastrading activities contribute 39%’.102 Assuming that the merged entity is able to increase costs for rivals downstream, the Commission will analyse to what extent such cost increases will give rise to price increases that result in customers switching to the downstream division of the merged entity. In its case law, the 99 101
102
100 Ibid., para. 295. Ibid., paras. 301–6. Case COMP/M.3868 – DONG/Elsam/Energi E2, Commission decision of 14 March 2006. Ibid., para. 363.
vertical mergers: input foreclosure
423
Commission has identified a number of factors that are relevant to this analysis such as the pass-through of costs to end-customers in the form of price increases, the elasticity of the end-demand and cross elasticity with respect to the merged entity, efficiencies and the size of the existing customer base of the merging parties. As explained above, the less important the input is to the final product, the more substantial the price increase needs to be to have an impact on prices of the relevant downstream products. Indeed, where the input only represents a minor part of the variable cost of downstream products, even a significant increase in the price of the input does not lead to an appreciable impact on variable costs downstream. Moreover, downstream rivals are unlikely to pass on the entire cost increase, especially when they face greater competition from the merged entity that may be in a position to offer lower prices due to efficiencies. The Commission made such a point in, for instance, Itema/Barcovision:103 indeed, most market participants have mentioned during the market investigation that the merger is likely to lead to a decrease in prices for Itema winders equipped with BarcoVision sensors. To continue competing with Itema, downstream competitors are therefore unlikely to pass on the entirety of their cost increase, which further limits the downstream effects of an input foreclosure scenario.104
In Nokia/Navteq,105 where the map databases accounted for only a small share of the mobile handsets’ wholesale prices, the Commission applied a scenario-analysis using different input price increases as well as pass-on rates, which showed that ‘under any reasonable own-price elasticity and diversion rate to the merged entity, such a small price increase would lead to very few additional sales for the merged firm’.106 Similarly, in TomTom/Tele Atlas,107 the Commission reached the same conclusion and noted: since map databases account on average for less than 10% of the PND wholesale price, map database prices would have to increase substantially to have an effect on downstream PND market prices and allow the merged entity to capture a significant amount of sales on the downstream market. Moreover, the impact of the foreclosure strategy depends on the
103 104 105 106 107
Case COMP/M.4874 – Itema/Barcovision, Commission decision of 4 August 2008. Ibid., para. 75. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. Ibid., para. 336. Case COMP/M.4854 – TomTom/Tele Atlas, Commission decision of 14 May 2008.
424
non-horizontal mergers extent to which TomTom’s competitors would pass on the map database price increase to end-consumers. For example, a 10% price increase of the map would only lead to a 0.5% price increase for the PND if the price of the map represents 10% of the price of a PND and PND manufacturers pass on 50% of the change in their cost.108
Even if all downstream rivals were to increase prices as a result of the foreclosure strategy, this might give rise to a limited increase in sales for the downstream division of the merged entity if customers would switch to other substitutable products in response to such a price rise. The Commission concluded in Arsenal/DSP109 that an input foreclosure strategy aimed at downstream mono-benzoate plasticiser producers was unlikely to be profitable since a large proportion of customers would prefer to replace mono-benzoate plasticisers by substitutable phthalate plasticisers rather than testing a new formula containing di-benzoate plasticisers.110 This meant that the merged entity’s downstream dibenzoate plasticisers business was unlikely to capture the necessary amount of sales to compensate for the upstream losses. In E.ON/MOL,111 the Commission highlighted the potential cost savings upstream that the merged entity would experience following non-pricing input foreclosure (i.e. lowering the quality of the service).112 The Commission also affirmed that economies of scale were contributing to increase the merged entity’s incentives to foreclose. In particular, the Commission noted: when the marginal cost of downstream gas supply is increasing, then it becomes progressively more costly for the integrated firm to replace its rivals’ sales with its own. Hence the integrated firm may find it profitable to engage in some nonprice discrimination, but not to raise its rivals’ costs so high that they exit the market. An extreme form of this occurs when the integrated firm has a capacity constraint on its downstream production. In contrast, in this case, it appears that there exist economies of scale in the retail supply of gas to small industrial and commercial customers. This is due to fixed costs in developing a sales network able to induce consumers under the regulated segment to switch to the open segment of the market.113
108 109 110 111 112
Ibid., para. 216. Case COMP/M.5153 – Arsenal/DSP, Commission decision of 9 January 2009. Ibid., paras. 297–300. Case COMP/M.3696 – E.ON/MOL, Commission decision of 21 December 2005. 113 Ibid., para. 430. Ibid., para. 431.
vertical mergers: input foreclosure
425
Moreover, the Commission stated that the little differentiation in the retail supply of gas contributed to consumers’ readiness to switch between downstream suppliers, and therefore gave rise in turn to a higher likelihood of success for an input foreclosure strategy.114 Low switching costs in the upstream market have the opposite effect. In TomTom/Tele Atlas,115 the Commission found that given that switching costs were surmountable, in the event of an increase of the input price by the merged entity, Tele Atlas (upstream) would lose a significant amount of sales to Navteq, which would hamper the ability of the merged entity to raise downstream rivals’ costs and limit the incentive to engage in the strategy.116 In Thales/Finmeccanica/Alcatel Alenia Space & Telespazio,117 the uncertainty on whether a TWT (input) foreclosure strategy would result in an increase of the merged entity’s likelihood of winning satellite bids (downstream) was considered to contribute to undermine the merged entity’s incentives to engage in such a strategy.118
3.2.3 Case study: Nokia/Navteq – assessment of incentive to engage in input foreclosure As part of its assessment of the Nokia/Navteq119 merger, the Commission carried out a detailed analysis of the merged entity’s incentive to engage in input foreclosure. Specifically, it considered the trade-off between profits lost in the upstream market for digital maps due to a reduction of input sales and the profit earned on the downstream market as a result of increases in the costs of rival mobile handsets producers.120 The Commission noted that the gross margin obtained on the sale of a mobile handset was much higher than the gross margin obtained on the sale of a map database. Whilst the margins themselves are redacted, the Commission’s decision suggests that the gross margins per handset 114 115 116 117
118 119 120
Case COMP/M.3696 – E.ON/MOL, Commission decision of 21 December 2005, para. 429. Case COMP/M.4854 – TomTom/Tele Atlas, Commission decision of 14 May 2008. Ibid., para. 219. Case COMP/M.4403 – Thales/Finmeccanica/Alcatel Alenia Space & Telespazio, Commission decision of 4 April 2007. Ibid., para. 307. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008. The Commission focused on assessing the incentive to foreclose rival producers of mobile handsets rather than the market for navigation applications on mobile handsets, in part due to data availability and in part because it considered that any incentive for the merged entity to engage in input foreclosure with regard to firms active in the market for navigation applications on mobile handsets would be weaker given Nokia’s more limited presence.
426
non-horizontal mergers
earned on the sale of digital maps to mobile handset manufacturers are many times smaller than the gross margins generated by the sale of a mobile handset.121 As explained in the example above, all else being equal, this would suggest that the critical increase in mobile handset demand required for foreclosure to be profitable might be relatively low: winning an incremental mobile handset sale could compensate for the loss of a number of digital maps sales. In Nokia/Navteq the question of which costs should be considered variable for the purposes of calculating gross margins was apparently uncontroversial. It should be noted, however, that in certain cases the assessment of which costs would be avoided upstream if the firm reduced sales as part of a foreclosure strategy and which costs would be incurred regardless of foreclosure is more complex. In some cases, it may be necessary to undertake a detailed examination of the firms’ cost structures, focusing particularly on the extent to which each cost item would be saved if the upstream firm reduced sales and to which each cost item would be increased if the downstream firm made additional sales. Having established that the critical increase in downstream demand for mobile handsets might be relatively low in Nokia/Navteq, the Commission’s analysis focused on whether the actual increase in downstream demand resulting from any foreclosure strategy would be sufficiently low to imply that foreclosure would not be profitable. The Commission identified a number of factors that suggested that the actual increase would indeed be low. First, the Commission noted that, since map databases account for a small proportion of mobile handset costs, map database prices would have to increase substantially to have an effect on mobile handset prices. Assuming that map databases represent 10% of mobile handset costs (the actual percentage is redacted) then a 10% increase in the price of map databases would increase mobile handset costs by 1%. If mobile handset manufacturers were to pass on 50% of this cost increase to consumers in the form of higher prices, the overall increase in downstream prices would be only 0.5%. The Commission concluded that under any reasonable assumption regarding own-price elasticity and diversion ratios between downstream firms and Nokia, such a small price increase would lead to very few additional sales for the merged entity.122 121 122
Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008, para. 334. Ibid., para. 336.
vertical mergers: input foreclosure
427
Second, navigation services are only one feature among many that determine customers’ handset choices (such as the handset’s music or video player). Faced with a degradation of quality or increase in price for digital maps, Nokia’s rivals could enhance other features of their handsets in order to win customers.123 Third, the Commission considered the role played by Garmin and, in particular, Garmin’s agreement with Samsung to offer a navigation solution on mobile handsets. It considered that this protection from foreclosure would limit the profits that Nokia could capture on the downstream market.124 In the context of our example above, if there were a third downstream firm (Firm 3) that were insulated from the effect of any foreclosure strategy, this would reduce the share of the benefits from the increase in Firm 2’s costs that are appropriated by Division 1. The Commission also noted a number of factors that it considered would imply that Navteq could lose a significant amount of upstream sales to Tele Atlas if it increased price or degraded map database quality. These factors included relatively limited switching costs and the view that Tele Atlas would continue to provide good-quality map databases to all mobile handset manufacturers in a non-discriminatory manner.125 It is arguably the case that these factors should have been considered as part of the Commission’s assessment of the merged entity’s ability to engage in foreclosure. If customers could switch easily to a product that was of equally high quality in response to a degradation of the quality of the product offered by the merged entity, then this would undermine the ability of the merged entity to affect the terms available on the market as a whole (with regard to quality) on which downstream rivals could purchase the input. In any event, the Commission’s consideration of these factors at the stage of assessing the merged entity’s incentive to foreclose highlights the blurred boundaries of the various stages of the competitive assessment of input foreclosure. The Commission’s analysis of the extent to which the merged entity could capture sales on the mobile handset market was supported by an econometric estimation of the demand for mobile handsets. Its estimate of downstream price elasticities implied, consistent with the above factors, that the merged entity would capture only a relatively limited volume of downstream sales following an increase in map database prices 123 124
Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008, para. 337. 125 Ibid., para. 338. Ibid., paras. 339 and 340.
428
non-horizontal mergers
charged for Nokia’s competitors. The Commission concluded that since map database prices represent a relatively minor proportion of the price of mobile handsets, the elasticity estimates it obtained coupled with conservative assumptions regarding the degree of cost pass-through by rivals, implied that a strategy of total foreclosure would not be profitable for the merged entity unless Tele Atlas increased its own upstream price by more than two hundred per cent (a scenario that it considered unlikely).126 The Commission also sensitivity-tested its analysis to take account of the possibility that map databases would represent a more important cost in the future, given that the price of other mobile components (such as hardware) might fall relative to map database prices. The Commission found that its conclusions were robust to variations in this assumption.127 Furthermore, the Commission noted the problem of commitment in its assessment. In particular, it pointed out that a strategy of total foreclosure that allowed Tele Atlas to increase prices by such a substantial amount might not be credible, as the merged entity would have an incentive to undercut any higher price offered by Tele Atlas.128 Since Tele Atlas would be able to predict this reaction from the merged entity, it would refrain from attempting to increase its prices in the first place. With regard to partial foreclosure, the Commission found that any price increase that would have a non-negligible impact on the downstream market would not be profitable for the merged entity as the downstream gains would not be sufficient to compensate upstream losses. This conclusion was robust to a range of alternative assumptions concerning the pass-through rate, elasticities and the relative importance of map databases as a cost.
3.3 Effect of input foreclosure Input foreclosure can be considered ‘anti-competitive foreclosure’ only in so far as it brings about higher prices in the downstream market as a result of a lessening of competition.
126 128
127 See ibid., para. 346. See ibid., para. 347. See ibid., para. 346. As discussed in our example, this incentive to undercut arises because it generates a significant increase in upstream profits but need not materially reduce the price at which downstream rivals can purchase the input.
vertical mergers: input foreclosure
429
The extent to which the increase in rivals’ costs caused by input foreclosure can be expected to give rise to higher downstream prices will depend significantly on whether the firms whose input costs have increased form a significant part of the market and in the aggregate represent a significant source of competition. Moreover, if the firms that are the subject of foreclosure are particularly close competitors for the downstream firm, this will increase the scope for the merged entity to increase price. On the other hand, if foreclosure only affects a small part of the downstream market (e.g. because certain downstream players rely less on the input whose price has increased), or there are potential entrants that are unaffected by the foreclosure strategy, then the merged entity is less likely to increase prices downstream. Any assessment of the effect of foreclosure should also take into account that over the longer term any increase in upstream prices may invite entry by new unintegrated firms at the upstream level, or may encourage downstream firms to vertically integrate upstream.129 Finally, as noted in Section 2.1, above, efficiencies can play an important role in the assessment of vertical mergers. A vertical merger, through the elimination of double marginalisation, may give rise to an incentive for the downstream division to lower its price relative to its pre-merger level. This may occur even if the merger simultaneously gives rise to input foreclosure that raises the costs and prices of rivals and that causes them to lose sales to the merged entity. Whilst competitors may be significantly harmed by the foreclosure strategy, the overall effect on consumers may be positive as a result of the lower prices offered by the merged entity. The Commission’s discussion of the effect of foreclosure in Nokia/ Navteq130 highlights the overlap between the various stages of assessment. The Commission noted that the same factors that explained the lack of any incentive to engage in partial foreclosure also implied that any 129
130
Note that one might consider that new entry at the upstream level that defeats any attempt to increase upstream prices should already have been taken into account at the stage of assessing the ability to foreclose. This reflects the somewhat blurred nature of the boundaries between the various stages of assessment. It might be that factors that affect the scope for the merged entity to raise rivals’ costs in the short term are considered at the ability to foreclose stage, whereas factors that affect the scope for the merged entity to do so over the longer term (e.g. new entry or vertical integration by downstream rivals) are considered at the stage of assessing the effect of foreclosure. Case COMP/M.4942 – Nokia/Navteq, Commission decision of 2 July 2008.
430
non-horizontal mergers
hypothetical foreclosure would not have a material effect in the downstream markets. In particular, it stated that ‘the low percentage of the map database in the mobile handset costs, the limited switching costs and the competition with Tele Atlas all tend to limit the price increase that could be imposed by Navteq on Nokia’s competitors.’131 This conclusion is interesting in so far as it suggests that the Commission essentially found that the merged entity would not have the ability to engage in partial foreclosure, despite the fact that earlier in its decision it had left open the question of whether the merged entity would have the ability to engage in foreclosure (see discussion at Section 3.1.3 above). Despite the lack of a finding of any anti-competitive effect arising from foreclosure, the Commission considered efficiencies that might result from the merger. It gave particular consideration to the potential for the merger to give rise to an elimination of double marginalisation given that the gross margins on map databases are high. The Commission set out the logic that post-merger the integrated company would realise that the true cost it faces for an additional map database is not 0–15% of the handset price but a small fraction of this amount. The Commission stated that if it is assumed that 30–60% of the cost decrease is passed on, the price of Nokia’s handsets would decrease by some small amount (less than 5%).132 The Commission also found that the elimination of double marginalisation would be merger specific to a large extent, since it did not observe contractual provisions in the market that served to eliminate double marginalisation.133
4.
Vertical mergers: customer foreclosure
Customer foreclosure refers to a situation in which a merger reduces the ability of the merged entity’s upstream competitors to make sales by restricting access to a significant part of the downstream customer base. To the extent that this restriction reduces the ability or incentive of upstream rivals to compete, this may lead to an increase in the costs of downstream rivals that in turn reduces the competitive pressure that the merged entity faces downstream. As with input foreclosure, if efficiency gains are not sufficient, the relaxation of this constraint 131
132
Ibid., para. 356. The Commission also took account of the ability of handset manufacturers and mobile network operators to source from Garmin, which was protected from price increases through its long-term contract with Navteq. See ibid., para. 361. 133 Ibid., para. 366. Ibid., para. 368.
vertical mergers: customer foreclosure Upstream firm
431
Upstream rivals Increasing upstream rivals’ cost
Reduced in purchases Efficiencies
Increasing downstream rivals’ cost Downstream firm
Figure 5.8
Reduced competition
Downstream rivals
Illustration of customer foreclosure concern
may lead to higher prices for final consumers. Figure 5.8 provides an illustration of these effects. Again, it must be stressed that ‘anti-competitive foreclosure’ is said to arise only in instances in which the result of foreclosure is a lessening of the effectiveness of competition faced by the merged entity that gives rise to consumer harm. Importantly, there is a close conceptual link between the logic of input and customer foreclosure theories. While the mechanisms in question differ, both forms of conduct represent means by which an integrated firm may seek to bring about a lessening in downstream competition via a worsening in the terms on which unintegrated downstream rivals obtain inputs. Customer foreclosure could thus be considered an indirect means by which integrated firms may bring about input foreclosure: rather than directly worsening the terms on which downstream rivals obtain inputs, customer foreclosure involves an integrated firm using its downstream position to worsen the terms on which independent upstream firms are able to supply downstream rivals. As a result, there is a high degree of overlap between the considerations relevant to the assessment of input and customer foreclosure concerns.134 134
In practice, however, the Commission has historically placed less emphasis on customer foreclosure concerns in merger assessment than input foreclosure theories. This may reflect the necessarily somewhat more contrived nature of customer foreclosure theories,
432
non-horizontal mergers
4.1 Ability to engage in customer foreclosure The first step in any assessment of whether a merger is likely to give rise to customer foreclosure is an analysis of the merged entity’s ability to engage in foreclosure. The ability to engage in customer foreclosure depends on the importance of the merged entity’s downstream business as a buyer of inputs sold by the merged entity’s upstream competitors. If as a result of reducing (or ceasing) purchases of inputs produced by third party suppliers the merged entity is able to restrict their access to market in a way that makes them less competitive in the upstream market, the merged entity is said to have the ability to engage in customer foreclosure. On the other hand, if the merged entity is not a significant buyer of the input compared to other buyers and as a result cannot influence the competitiveness of third party upstream competitors, the merged entity cannot be said to have the ability to engage in customer foreclosure. In practice, customer foreclosure may take the form of the downstream division of the merged entity refusing to source from upstream rivals, reducing its purchases from upstream rivals or only purchasing from third parties at terms that are less favourable for them. A key question in assessing the merged entity’s ability to foreclose is whether there are alternative sources of demand for the merged entity’s upstream rivals. If the pre-merger purchases of the downstream division from third party upstream suppliers are relatively small compared to the total sales made by those firms, then it is unlikely that even a complete refusal to source would have much of an impact on the ability of those upstream suppliers to continue to compete effectively. On the other hand, if the downstream division’s purchases from third parties represent a substantial part of their overall sales the withdrawal of those purchases could under some circumstances adversely affect their ability to compete. Consequently, as recognised by the EC Non-horizontal Merger Guidelines, market power in the downstream market is a necessary condition for customer foreclosure to arise.135 The extent to which a reduction in purchases by the merged entity will affect the competitiveness of third party upstream suppliers, and thus the prices faced by unintegrated buyers of the input, will typically depend on the cost conditions of those suppliers. There are two mechanisms by
135
which involve an additional step in the chain of logic before anti-competitive effect can be established. EC Non-horizontal Merger Guidelines, para. 61.
vertical mergers: customer foreclosure
433
which a reduction in sales could impair an upstream supplier’s ability to compete. First, and most obviously, upstream rivals could simply be forced to exit due to a lack of sales rendering their activities unviable. Second, where an upstream supplier’s unit production costs decline with the level of output, a reduction in sales will result in an increase in its marginal costs, thereby reducing its ability to compete on price. Thus, where the merged entity’s demand is an important factor in enabling a third party upstream supplier to achieve scale economies that ensure that it is attractive to remain in the market,136 or where the inability to serve the merged entity implies that unintegrated downstream firms can only be supplied at a higher incremental cost, this may imply that the ability of unintegrated firms to offer effective competition to the merged entity’s upstream division is reduced by the foreclosure strategy.137 This can lead to an increase in the input prices faced by rivals to the merged entity’s downstream division. In addition, if the reduction of expected revenues resulting from customer foreclosure is sufficient to deter upstream rivals from investing in R&D that might have reduced their incremental costs of supply or increased the quality of their products, this may adversely affect their ability to compete with the upstream firm over the longer term relative to the situation in which foreclosure had not taken place. Alternatively, even where a merged firm is able to withdraw significant purchases from third parties, this may not result in foreclosure if it does not affect the attractiveness to third parties of remaining in the market, nor their cost of making incremental sales to unintegrated downstream firms. This could be the case if many of the costs that would have been incurred in serving the merged entity’s downstream division can be avoided over the relatively short term, so that costs fall in line with any reduction in output, such that rivals’ incentive to remain active in the market is not reduced. Similarly, if incremental costs are unrelated to
136
137
Equivalently, where the merged entity’s demand is an important factor in enabling a third party potential entrant to the upstream market to achieve scale economies that ensure that it is attractive to enter the market. Note, the merged entity might be a substantial buyer in a particular market in which the unintegrated upstream supplier sells, but a relatively small buyer in relation to the entire base of sales over which the upstream supplier achieves economies of scale. These economies of scale could derive, for example, from sales across a range of different geographic markets. Even if the merged entity has a significant position in one of those markets, its purchases may not be a material determinant of the upstream firm’s costs of supply.
434
non-horizontal mergers
volumes sold then, as long as foreclosure does not induce exit, upstream rivals will not become any less competitive.138 In order for customer foreclosure to give rise to a significant increase in the costs of downstream rivals, it is necessary that the input supplied by the upstream firms that are the target of foreclosure must in some sense be important for those downstream rivals. If the cost increase caused by customer foreclosure relates to an input that represents a relatively trivial part of the costs of supply of downstream rivals, then the merged entity is likely to be unable to affect the costs of downstream rivals through customer foreclosure.139 In any particular case, the ability of the merged entity to foreclose may depend on a range of factors specific to the case at hand. For example, there may be factors that make it difficult for the downstream division of the merged entity to suddenly switch its demand away from competitors of the upstream division. This may be the case if there are barriers to switching large volumes of demand. In addition, as discussed further below, the merged entity’s upstream rivals may engage in counterstrategies (or even counter-mergers) that undermine the ability of the merged firm to engage in customer foreclosure.
4.1.1 Ability to engage in customer foreclosure: an example The following is a highly stylised example of how a firm that merges with a downstream customer, and then refuses to source from its upstream rivals, can increase its market power upstream and thereby raise the input costs of rivals to its downstream division. In this example, the goal of customer foreclosure is to create market power upstream by preventing the entry of an international potential entrant. Suppose Firm A and Firm B are producers of a component, which is an input to the production of a gadget produced by Firm 1 and Firm 2. Firm A is relatively inefficient with a marginal cost of production of 140. Firm B, on the other hand, is relatively efficient with a marginal cost of only 50. Importantly, however, whilst Firm A faces no fixed costs, Firm B faces a fixed cost of 10,000 if it is to serve Firms 1 and 2. This is because 138
139
It should be noted in this regard that increases in average costs arising from the spreading of fixed costs over a smaller sales base will not impact the pricing incentives of third party rivals. In practice, the importance of the cost of the input as a proportion of downstream rivals’ total costs may be taken into account at the stage of assessing ability to foreclose (as it is in the case of input foreclosure) or at a later stage of the overall assessment.
vertical mergers: customer foreclosure
435
Firm B currently produces components for a different geographic market in which the gadgets are somewhat different to those produced by Firms 1 and 2. The fixed cost faced by Firm B represents a one-off investment in R&D required to make its components compatible with gadgets produced by Firms 1 and 2. We assume that having made this R&D investment the components produced by Firms A and B are homogeneous. Suppose that the downstream market for gadgets is of a fixed size of 200 units and that one component is required to produce each gadget. To understand how a merger between Firm A and Firm 1 could give rise to customer foreclosure resulting in higher input prices for the merged entity’s downstream competitor, Firm 2, we first consider market outcomes absent any merger. First, we calculate the price that enables Firm B to break-even if it were to enter and then supply the entire market (200 units). This is the price that ensures that revenues are equal to costs. The total cost of entering and supplying 200 units is equal to the unit cost (50) multiplied by the number of units supplied (200) plus the one-off fixed cost (10,000), which is equal to 20,000. Given that supplying the market results in 200 gadget sales, the price that allows Firm B to break-even with this strategy is 100 (since 100 * 200 ¼ 20,000). Importantly, this price is below the marginal cost of Firm A. Absent any merger, we might therefore expect Firm B to offer to supply the entire market at a price of 140. This strategy is profitable for Firm B because it serves the market at a price higher than the break-even price of 100. Given that Firm A has a marginal cost of 140 (and therefore cannot profitably offer a lower price than this), Firms 1 and 2 can do no better than to accept Firm B’s offer. This situation is illustrated in Figure 5.9. We now suppose that Firm A merges with Firm 1 to create Firm A1.140 We suppose that Division 1 engages in a strategy of only sourcing from Division A (at cost), regardless of the price that Firm B offers. This implies that Firm B can now only bid for the business of Firm 2 and must make its entry decision based on whether the revenues from serving Firm 2 will cover its total costs (including the fixed R&D cost of 10,000). Suppose (for the time being) that Firm 2’s demand amounts to half of the market demand (i.e. 100 components). We now calculate the price that enables Firm B to break-even if it were to win Firm 2’s business. 140
As above, we refer to the upstream division of the merged entity, Firm A1, as Division A and to the downstream division as Division 1.
436
non-horizontal mergers A
B
P = 140
1
Figure 5.9
P = 140
2
Illustration of pre-merger scenario
The total cost of entering and supplying 100 units is equal to the unit cost (50) multiplied by the number of units supplied (100) plus the one-off fixed cost (10,000), which is equal to 15,000. The price that allows Firm B to cover this cost from 100 gadget sales is 150 (since 100 * 150 ¼ 15,000). Therefore, Firm B will not find it profitable to agree to supply Firm 2 at any price lower than 150. This enables Division A to offer to supply Firm 2 at a price of 150. This strategy is profitable for Firm A because this price is above its marginal cost of supply of 140. Moreover, given that Firm B cannot profitably offer to serve Firm 2 at a price lower than this, Firm 2 can do no better than to accept Firm A’s offer. This situation is illustrated in Figure 5.10. In this example, customer foreclosure via Division 1 withdrawing its purchases from Firm B has thus raised the price at which the unintegrated firm purchases components from 140 to 150.141 We now consider a number of factors that affect the extent to which Division 1’s refusal to source from Firm B gives rise to customer foreclosure that increases Firm 2’s costs. First, the size of the contestable downstream demand that is unaffected by the merged entity’s refusal to source is a crucial determinant of the merged entity’s ability to foreclose. If instead of having a demand of
141
To the extent that Firm 2’s demand for components is in fact lower than 100 at the higher price of 150 (because it loses some gadget sales to Firm 1 as a result of passing through some of this cost increase in the form of higher gadget prices), then the price that enables Firm B to break-even by serving Firm 2 may be somewhat higher than 150. For example, at price of 150, Firm 2’s demand for components may now only be 95 units. Firm B can only profitably supply 95 components at a price of 155.26, which allows Firm A some further scope to increase its price.
vertical mergers: customer foreclosure
A
P = 140
B
P = 150
1
Figure 5.10
437
2
Illustration of post-merger scenario with customer foreclosure
100 units, Firm 2 had a demand of 110 units, the merged entity would no longer have the ability, through refusing to source from Firm B, to charge a price significantly higher than its marginal cost of 140 to Firm 2. This is because the price that would allow Firm B to enter and profitably serve Firm 2’s demand of 110 is now around 141. At a price of 141, Firm B’s revenues from serving Firm 2 would be 15,510 (equal to 141 * 110 units), which is just higher than its total cost of 15,500 (equal to 50 * 110 + 10,000). Therefore, any attempt to offer a price of 141 to Firm 2 would be undermined by entry by Firm B. Second, the size of the fixed cost of entry faced by Firm B in this example is an important determinant of the merged entity’s ability to engage in foreclosure. Assuming again that Firm 2’s demand is for 100 units, if the fixed cost of entry were only 9,000 rather than 10,000, this would eliminate the scope for Division A to increase the input price faced by Firm 2 above its marginal cost of 140. This is because if the fixed cost of entry were only 9,000 this would imply that the price at which Firm B can just break-even and serve Firm 2 is 140. To see this, consider that at the price of 140, its revenues (14,000 ¼ 100 * 140) are just sufficient to enable it to cover its costs (14,000 ¼ 100 * 50 + 9,000). Therefore, given the lower fixed cost faced by Firm B of 9,000, Firm A would have no ability to raise the price above its marginal cost of 140. Any attempt to do so would allow Firm B to enter and serve Firm 2 at a lower price.
438
non-horizontal mergers
Third, the effect of any increase in the price of the input on the cost of downstream firms will depend on the importance of the component as a proportion of total costs. If components represent 100% of gadget producers’ marginal costs, then an increase in component prices due to customer foreclosure would likely undermine Firm 2’s ability to constrain the merged entity. Alternatively, if components only represent 10% of Firm 2’s marginal cost, then an increase in the price of the component of 7% (i.e. from 140 to 150) would only cause its marginal cost of producing gadgets to increase by 0.7%.142 It should be noted that the example above is highly stylised. The results are driven by the specific assumptions and in particular by the fact that average costs of supply for the potential entrant vary significantly with the volume it sells due to the large fixed cost of entry it faces. Hence, its decision to enter is heavily dependent on the volume of demand that is contestable. The example is also somewhat limited by the assumption that the demand of Firm 2 is given. If Firm 2 were able to expand sales substantially by offering a lower downstream price (both through taking sales away from Firm 1 and through expanding the market) then this could enable Firm B to counter the merged entity’s foreclosure strategy by selling components at a low price to Firm 2 that enables it to expand its volume of sales. Through expanding the contestable base in this way, Firm B reduces the minimum price at which it can profitably enter. It could be that a profitable counterstrategy for Firm B exists that involves helping Firm 2 achieve a large sales volume by offering it a low input price. For example, if, by undercutting Firm 1, Firm 2 could achieve sales of 200, then the price that would allow Firm B to break-even on sales to Firm 2 would fall to only 100. The counterstrategy of offering Firm 2 a price of 100 could in this case defeat any ability of the merged entity to engage in customer foreclosure. Firm B could even merge with Firm 2 as part of its entry strategy to ensure it priced aggressively downstream and secured a demand base sufficient to render its R&D investment worthwhile.143 142
143
As noted above, the importance of the cost of the input as a proportion of downstream rivals’ total costs may be taken into account at the stage of assessing ability to foreclose (as it normally is in the case of input foreclosure) or at a later stage of the overall assessment. In practice, the ability of a merged entity to engage in customer foreclosure depends on a rich set of considerations. These include the ability of upstream rivals to engage in counterstrategies, which may include vertical mergers of their own.
vertical mergers: customer foreclosure
439
4.2 Incentive to engage in customer foreclosure A vertical merger alters the incentives of the downstream division because it will now take account of the impact of its purchasing decisions on its upstream division. This may give rise to an incentive to engage in customer foreclosure. Whether a firm has an incentive to engage in customer foreclosure depends on whether the benefits of customer foreclosure outweigh the costs to the merged entity as a whole. The costs associated with customer foreclosure stem from the fact that it involves the downstream division reducing or ceasing its purchases from certain upstream suppliers, which, absent the foreclosure strategy, would have been purchases it preferred to undertake. For example, the foreclosure strategy may involve the downstream division switching purchases from a more efficient upstream supplier to the upstream division of the merged entity. Therefore, the downstream division may have achieved lower input costs if it sourced from third party upstream firms rather than its own upstream division.144 Alternatively, it may have procured a higher quality input absent the foreclosure strategy. The benefit from engaging in customer foreclosure may stem from two factors. First, if the customer foreclosure strategy makes upstream rivals less competitive, this may enable the merged entity’s upstream division to increase its margins on sales to rivals of its downstream division (or increase sales to downstream firms). Second, if the input costs of downstream rivals are increased as a result of the foreclosure strategy, this may relax the competitive constraints faced by the downstream division, enabling it to increase price and/or increase its volumes. A number of factors determine whether the overall impact on profits will be positive leading to an incentive to foreclose. As with our discussion of the incentive to engage in input foreclosure, we consider the incentive to engage in customer foreclosure by reference to two concepts: the critical increase in downstream demand required to make the foreclosure strategy profitable; and the actual increase in downstream demand that occurs as a result of the customer foreclosure strategy. As discussed in more detail in the following subsection, the critical increase in downstream demand required to make foreclosure profitable will
144
EC Non-horizontal Merger Guidelines, para. 69.
440
non-horizontal mergers
depend on the relative margins in upstream and downstream activities both in the presence of and absence of foreclosure.145 The actual increase in downstream demand that results from the foreclosure strategy will depend on the effect of the foreclosure strategy on the costs of downstream rivals;146 the extent to which any given cost increase is passed on to consumers by downstream rivals in the form of higher prices; and the extent to which those downstream consumers react by substituting to the products of the merged entity’s downstream division.147
4.2.1 Incentive to engage in customer foreclosure: an example In the example set out in Section 4.1.1, the merged entity had the ability to refuse to source from B and thereby to increase its rival’s input price from 140 to 150. We now use a slightly modified version of the same example to illustrate the assessment of whether a vertically integrated firm might have an incentive to engage in customer foreclosure. Suppose that Firm B offers to supply the entire market at a price of 100, for instance due to the presence of a third potential upstream supplier, Firm C, with the same cost structure as Firm B. A price of 100 is just profitable for B in the event that it sells 200 units since revenue (100 * 200 ¼ 20,000) equals cost (200 * 50 + 10,000 ¼ 20,000). Suppose further that in the event that Firm 1 refuses this offer, Firm B offers to supply Firm 2 at a price of 150. This price is just profitable for Firm B in the event that it sells 100 units since revenue (150 * 100 ¼ 15,000) equals cost (100 * 50 + 10,000 ¼ 15,000). We assume that the gadgets produced by Firms 1 and 2 are differentiated, but that, as before, the components produced by Firm A and Firm B are homogeneous. We assume that if the merged entity accepts Firm B’s offer, both Firm 1 and Firm 2 have a unit cost of 100 and sell 100 gadgets each at a price of 200. 145
146
147
The critical increase in demand for the downstream division could be zero if the increase in margin upstream due to the foreclosure strategy is equal to the reduction in margin downstream resulting from sourcing the input internally at a higher cost to that available for sourcing from third parties. This in turn depends on the factors discussed in Section 4.1 above, such as the importance of upstream scale economies and the importance of the input as a cost factor. The profitability of this strategy is also contingent on the downstream division having the necessary spare capacity to meet any increase in demand the foreclosure strategy generates.
vertical mergers: customer foreclosure Price, Cost
150 140
Price, Cost 200
Upstream
441
Downstream
6,000 1,000
100
q
100
q
Figure 5.11 Upstream and downstream profits of merged entity under customer foreclosure absent changes in prices and output downstream
First we consider Firm A1’s profit if it accepts Firm B’s offer to supply Division 1. Both Division 1 and Firm 2 accept Firm B’s offer to supply the entire market at a price of 100. In this case, the profits of Division A are zero. The profits of Division 1 on the other hand are equal to the price it charges for gadgets (200) less the marginal cost (100) multiplied by the number of gadgets it sells (100), which equals 10,000. Thus, in the event that it accepts B’s offer, Firm A1’s profits are 10,000. We now consider Firm A1’s profits if it refuses Firm B’s offer to supply Division 1. In this case, as we noted in Section 4.1.1, Division A will be able to offer a price of 150 to supply Firm 2.148 The upstream division’s profits are therefore equal to the unit price (150) less the marginal cost (140) multiplied by the number of components sold to Firm 2 (100), which equal 1,000. The downstream division is supplied by the upstream division at marginal cost (140). For the time being, we assume that Division 1 will price at 200 even at this higher cost and even though Firm 2’s costs have increased (although we will see the implications of relaxing this assumption later). Division 1’s profits are equal to the price it charges for gadgets (200) less the marginal cost (140) multiplied by the number of gadgets sold to consumers (100), which equal 6,000. The total profit of the merged entity Firm A1 in the event that it refuses to source from Firm B is therefore equal to 7,000 (as illustrated in Figure 5.11). It is therefore apparent that absent any change in the prices and outputs of the downstream Firms 1 and 2, the merged entity can obtain
148
Given that Firm B cannot profitably offer to serve Firm 2 at a price lower than this, Firm 2 can do no better than to accept Firm A’s offer.
442
non-horizontal mergers
higher profits by accepting the offer of Firm B to supply its downstream division. The cost of refusing to source components from Firm B arises from the fact that it involves forgoing procurement of the component at a price that is lower than its own marginal cost of supply. However, given that the foreclosure strategy has increased the input cost of Firm 2 from 140 to 150, this might be expected to cause it to increase the price it charges final consumers for gadgets. To the extent that this increase in input price causes some consumers to switch from Firm 2’s gadgets to Division 1’s gadgets, this increase in downstream demand may be sufficient to render the foreclosure strategy profitable overall. It is now possible to calculate, holding prices in the downstream market constant at 200, how many units would need to switch from Firm 2 to Division 1 in order to justify the cost of refusing to source from Firm B. In other words, we can calculate the critical increase in Division 1’s demand that is required to make this strategy profitable.149 We also assume for the time being that Division 1’s downstream price does not change as a result of the foreclosure strategy so that the benefits of foreclosure are down purely to an increase in its volume.150 First recall that the shortfall in profit resulting from the refusal to source case compared to the alternative was 3,000 (equal to 10,000 – 7,000). Note that every unit that shifts from Firm 2 to Division 1 as a result of the higher downstream price charged by Firm 2 adds 50 to the merged entity’s profits. This is because each incremental unit sold allows the downstream division to earn a margin of 60 (equal to 200 – 140) but at the same time each incremental unit switched also represents a lost sale for the upstream division to Firm 2, on which it would have earned a margin of 10 (equal to 150 – 140). The net effect on profit of a unit switched is therefore 50. It can then be seen that the number of units that must be switched in order to increase the merged entity’s profit from 7,000 to 10,000 (the profit that can be earned by sourcing from Firm B) is 60, since a net margin of 50 per unit multiplied by 60 units equals 3,000. The switch of 60 units reduces the upstream division’s profit from 1,000
149
150
In what follows we abstract away from the issue that any decrease in Firm 2’s demand will mean that Firm B’s price offer of 150 to serve Firm 2 will no longer be profitable, and that this might enable Division A to further increase the price of the component to Firm 2 above 150. We will later come on to see how an increase in Division 1’s price due to the relaxation of competition downstream affects the critical increase in downstream demand.
vertical mergers: customer foreclosure Upstream
Price, Cost
150 140
Price, Cost 200
443
Downstream
9,600 400
40
100
q
100
160 q
Figure 5.12 Upstream and downstream profits of merged entity under customer foreclosure assuming 60 units switched to downstream division
to 400 (since it now sells 40 rather than 100 units at a margin of 10 per unit). But it increases the downstream division’s profit from 6,000 to 9,600 (since it now sells 160 rather than 100 units at a margin of 60 per unit). The combined entity therefore obtains profits of 10,000 (as illustrated in Figure 5.12). The critical increase of 60 units calculated above depends on the upstream margin lost on incremental sales to Firm 2, the margin earned on downstream sales given the refusal to source and the margin that would have been earned on downstream sales had the merged entity accepted Firm B’s offer to supply Division 1 at a price below Division A’s marginal cost. The critical increase required to make the strategy profitable would be smaller than 60 units if: • The margins earned on upstream sales to Firm 2 given foreclosure were higher. Suppose that as a result of foreclosure, the upstream division could raise the price it charges Firm 2 to 160 rather than 150.151 This would imply that the margin earned on upstream sales is now 20 (160 – 140) and that profits from the sale of 100 units to Firm 2 now earn the upstream division 2,000. In the absence of any increase in downstream volume, the profits of the downstream division would be 6,000 (since it makes 100 gadget sales at a margin of 60). The total profit of the merged entity given foreclosure would therefore be 8,000 in the absence of any increase in downstream demand. Given that, as
151
This might be possible, for example, if the fixed cost of entry were 11,000, rather than 10,000. A fixed cost of entry of 11,000 would imply that the price required to break-even on sales of 1,000 units would be 160, since at this price revenues (100 * 160) would equal cost (11,000 + 50 * 100) for Firm B.
444
non-horizontal mergers
calculated previously, the merged entity will generate profits of 10,000 if it instead accepts Firm B’s offer, the net effect on profits resulting from taking volumes from Firm 2 must amount to at least 2,000 in order for the foreclosure strategy to be profitable. Every unit that shifts from Firm 2 to Division 1 as a result of the higher downstream price charged by Firm 2 adds 40 to the merged entity’s profits.152 Therefore the critical increase in demand is now 50 (since 40 * 50 ¼ 2,000).153 • The margins earned on downstream sales to consumers given foreclosure were higher. Suppose that as a result of the relaxation of competition downstream caused by the foreclosure strategy, Division 1 were able to increase its gadget price from 200 to 210.154 This would imply that the margin earned on downstream sales is now 70 (210 – 140) and that, absent any shift in purchasing by consumers, downstream profits from the sale of 100 units are 7,000. The upstream division makes profits of 1,000 from the sale of 100 units to Firm 2 at a margin of 10 (150 – 140). The total profit of the merged entity given foreclosure would therefore be 8,000. Given that the merged entity will generate profits of 10,000 if it instead accepts Firm B’s offer, the net effect on profits resulting from taking volumes from Firm 2 must amount to at least 2,000 (10,000 – 8,000) if foreclosure is to be profitable. Every unit that shifts from Firm 2 to Division 1 as a result of the higher downstream price charged by Firm 2 adds 60 to the merged entity’s profits.155 Therefore, the critical increase in demand is now 33.33 (since 60 * 33.33 ¼ 2,000).156 152
153
154
155
156
This is because each incremental unit sold allows the downstream division to earn a margin of 60 (equal to 200 – 140) but at the same time each incremental unit switched also represents a lost sale for the upstream division to Firm 2, on which it would have earned a margin of 20 (equal to 160 – 140). The net effect on profit of a unit switched is therefore 40. Note that if the upstream price could have been raised to 180 rather than 150, then the critical increase would be zero. The profit from foreclosure would equal the profit from accepting Firm 2’s offer even if there were no shift of demand downstream to the merged entity. Note that the downstream price might also be expected to be higher than 200 due to the fact that the downstream firm’s marginal cost is higher under the foreclosure strategy (140) than under the absence of foreclosure (100). This is because each incremental unit sold allows the downstream division to earn a margin of 70 (equal to 210 – 140) but at the same time each incremental unit switched also represents a lost sale for the upstream division to Firm 2, on which it would have earned a margin of 10 (equal to 150 – 140). The net effect on profit of a unit switched is therefore 60. Note that if the downstream price could have been raised to 230 rather than 210, then the critical increase would be zero. The profit from foreclosure would equal the profit
vertical mergers: customer foreclosure
445
• The margins earned on downstream sales to consumers absent foreclosure were lower. Suppose that instead of making an offer to supply the entire market at a price of 100, Firm B offered to do so at a price of 110. This would imply that absent foreclosure, the merged entity’s downstream margin would now be 90 rather than 100, and it would earn profits of 9,000 from 100 gadget sales. Given that profits for the merged entity from engaging in foreclosure are 7,000 absent any downstream demand shift, the net effect on profits resulting from taking volumes from Firm 2 must amount to at least 2,000 (9,000 – 7,000) if foreclosure is to be profitable. Every unit that shifts from Firm 2 to Division 1 as a result of the higher downstream price charged by Firm 2 adds 50 to the merged entity’s profits.157 Therefore, the critical increase in demand is now 40 (since 50 * 40 ¼ 2,000).158 Whether the customer foreclosure strategy is profitable depends on whether the actual increase in Division 1’s sales is greater than the critical value. The extent to which the actual increase in Division 1’s demand will exceed the critical value will depend on: • the extent to which the increase in Firm 2’s marginal cost of 100 to 150 is passed on to end-consumers of gadgets; and • the extent to which consumers of gadgets faced with an increase in Firm 2’s price switch to gadgets supplied by Division 1 (i.e. the degree of substitutability between Firms 1 and 2’s gadgets). The higher the proportion of the component cost increase that is passed on to consumers in the form of higher gadget prices, the lower need be the substitutability between the gadgets produced by Firm 2 and Division 1 in order for the actual increase in Division 1’s demand to exceed the critical amount. Conversely, the higher the substitutability between the downstream products, the lower need be the proportion of the cost increase that is
157
158
from accepting Firm 2’s offer even if there were no shift of demand downstream to the merged entity. This is because each incremental unit sold allows the downstream division to earn a margin of 60 (equal to 200 – 140) but at the same time each incremental unit switched also represents a lost sale for the upstream division to Firm 2, on which it would have earned a margin of 10 (equal to 150 – 140). The net effect on profit of a unit switched is therefore 50. Note that if Firm B offered a price of 130 rather than 100, then the critical increase would be zero. The profit from foreclosure would equal the profit from accepting Firm 2’s offer even if there were no shift of demand downstream to the merged entity.
446
non-horizontal mergers
passed on in order for the actual demand increase to exceed the critical demand increase.159 We noted above that if customer foreclosure enabled Division 1 to increase its margin, then this would reduce the critical demand increase required to make the foreclosure strategy profitable. It should be noted that any increase in Division 1’s price would also reduce the actual increase in demand it obtains due to the foreclosure strategy compared to the situation in which it held its price constant.160 As with the assessment of the incentive to engage in input foreclosure, it greatly simplifies the analysis of both the critical increase in demand and actual increase in demand if pre-merger downstream margins are used as a proxy for post-merger downstream margins. By making this assumption, it is often possible to gain significant insights into the profitability of a hypothesised foreclosure strategy for the purpose of assessing a firm’s incentive to engage in foreclosure.
4.3 Effect of customer foreclosure As with input foreclosure, customer foreclosure is only labelled ‘anticompetitive foreclosure’ if it results in higher prices in the downstream market as a result of a lessening of competition. Customer foreclosure can only give rise to harm to consumers if, by reducing the sales of rival upstream firms, it decreases competitiveness upstream causing the costs of downstream rivals that procure the input to increase and this in turn leads to higher prices in the downstream market. The extent to which upstream competitiveness is affected by the foreclosure strategy will depend significantly on the factors discussed under the assessment of the ability to engage in customer foreclosure such as the importance of scale economies and the size of the contestable downstream demand (i.e. demand that is unaffected by the foreclosure strategy). It will also depend on the extent to which the upstream firms affected by the foreclosure strategy represent an important source of 159
160
It should be noted that these factors are themselves interrelated since the extent to which an increase in Firm 2’s costs will lead it to increase price will partly depend on the extent to which any price increase will cause it to lose sales to Division 1. This is because some consumers of Firm 2’s gadget that might otherwise have switched following Firm 2’s price increase may in the face of higher prices for the merged entity’s gadgets remain with Firm 2 rather than switching.
vertical mergers: customer foreclosure
447
competition in the upstream market. If the foreclosure strategy only affects a small part of the upstream market (e.g. because certain important upstream players achieve minimum efficient scale through sales they make in other geographic markets), then the foreclosure strategy is less likely to reduce competition upstream. Even if the customer foreclosure strategy causes a reduction in competition upstream, and thereby increases the prices faced by certain downstream rivals, this does not imply that it will give rise to harm to consumers. The extent to which an increase in downstream rivals’ input costs caused by customer foreclosure can be expected to give rise to higher downstream prices will depend significantly on whether the firms whose input costs have increased form a significant part of the market and in the aggregate represent a significant source of competition. Moreover, if the firms that are affected by the reduction in upstream competition are particularly close competitors for the downstream firm, this will increase the scope for the merged entity to increase price. On the other hand, if foreclosure only affects a small part of the downstream market (e.g. because certain downstream players rely less on the input whose price has increased), or there are potential entrants that are unaffected by the foreclosure strategy, then the merged entity is less likely to increase prices downstream. Any assessment of the effect of foreclosure should also take into account that over the longer term any increase in upstream prices may invite entry by new unintegrated firms at the upstream level. As discussed in Section 4.1, upstream firms may be able to counter any customer foreclosure strategy by merging with a downstream firm and pricing aggressively to expand the base of sales over which it can achieve scale economies.161 Finally, any assessment of a vertical merger must take into account the potentially important role played by efficiencies (see Section 2.1). Through the elimination of double marginalisation, a vertical merger will often give rise to an incentive for the downstream division to lower
161
As with assessment of the effect of input foreclosure, the boundaries between the various stages of assessment are blurred. It might be that factors that affect the scope for the merged entity to raise upstream rivals’ costs in the short term are considered at the ability to foreclose stage, whereas factors that affect the scope for the merged entity to do so over the longer term (e.g. new entry or countervailing mergers with downstream firms) are considered at the stage of assessing the effect of foreclosure.
448
non-horizontal mergers
its price relative to its pre-merger level. This may occur even if the merger simultaneously gives rise to customer foreclosure that raises the costs and prices of downstream rivals and that causes them to lose sales to the merged entity. While competitors, both upstream and downstream, may be significantly harmed by the foreclosure strategy, the overall effect on consumers may be positive as a result of the lower prices offered by the merged entity.
4.4 Customer foreclosure: EU case law Although customer foreclosure is not often raised in vertical merger cases, the Commission identified this type of concern in a number of energy mergers that combined the incumbent suppliers in vertically related markets. For instance, in ENI/EDP/GDP,162 the Commission advanced customer foreclosure concerns in the wholesale gas market. The proposed transaction, which was ultimately prohibited after a Phase II investigation, would have brought under common ownership the monopoly supplier of wholesale gas in Portugal and the only supplier active in the downstream markets for gas-fired power electricity generation and retail distribution of gas and electricity. The Commission concluded that the merger as notified would have foreclosed potential rivals from gaining access to sufficient demand for gas to make entry viable in the upstream wholesale gas market. According to the Commission, the proposed transaction would have therefore reinforced GDP’s dominant position upstream.163 Similarly, in DONG/Elsam/Energi E2,164 the Commission assessed the likelihood of customer foreclosure in the wholesale gas market. DONG was a quasi-monopolistic company upstream, while Elsam and E2, the Danish electricity generation incumbents in West Denmark and East Denmark respectively, were by far the most important Danish purchasers of natural gas. After analysing the long-term contracts between each of the downstream parties and DONG, the Commission concluded that a foreclosure effect would likely occur for short-term volumes, and would cover all Elsam and E2 sales following the long-term contracts’ expiry. The Commission concluded that this customer foreclosure effect would 162 163 164
Case COMP/M.3440 – ENI/EDP/GDP, Commission decision of 9 December 2004. See ibid., Section V.B.2. Case COMP/M.3868 – DONG/Elsam/Energi E2, Commission decision of 14 March 2006.
vertical mergers: customer foreclosure
449
constitute a significant strengthening of DONG’s dominant position upstream, and that this would create further disincentives for potential entrants. In particular, the decision reads: ‘the foreclosure of this demand after the merger will make it more difficult for competitors to DONG to enter the Danish natural gas sector, whether as wholesale suppliers or as suppliers of final customers, thereby raising barriers to entry to all of these markets.’165 In Syngenta/Monsanto’s Sunflower Seed Business,166 competitors and potential entrants in the markets for flower seed treatment (upstream) indicated that the merger would prevent them from reaching the necessary customer base to compete effectively. In particular, customer foreclosure concerns were put forward for two different upstream markets: the market for sunflower seed treatment fungicides and the market for sunflower seed treatment insecticides. Ultimately, the Commission concluded that the merged entity would not have the ability to foreclose current competitors or potential entrants in these markets, since the proposed transaction would lead to a combined market share of [40–50%] in the market for the commercialisation of sunflower seed (downstream), leaving at least [50–60%] of the demand available for other seed treatment producers (upstream).167 In a number of other cases, the Commission was able to dismiss customer foreclosure concerns without the need to undertake a detailed empirical analysis. For example, in Itema/Barcovision,168 the Commission readily excluded the possibility that the transaction could give rise to customer foreclosure by pointing out that: ‘Uster [Barcovision’s main rival]’s sales of winder sensors to Itema represent only [10–20]% of its total sales and less than 10% of the total winder sensors market. Should Itema stop buying sensors from Uster, this would have limited effects on the upstream markets, making a customer foreclosure strategy clearly unprofitable.’169 Finally, the Commission considered the potential for anti-competitive customer foreclosure during its investigation of the acquisition of BST, a company active in the production of technical fabrics, by WLR, a private 165
166
167 168 169
Case COMP/M.3868 – DONG/Elsam/Energi E2, Commission decision of 14 March 2006, para. 504. Case COMP/M.5675 – Syngenta/Monsanto’s Sunflower Seed Business, Commission decision of 17 November 2010. See ibid., Sections IX.4.1.2, IX.4.2.2 and IX.4.2.4. Case COMP/M.4874 – Itema/Barcovision, Commission decision of 4 August 2008. See ibid., para. 91.
450
non-horizontal mergers
equity firm that owned Safety Components International ‘SCI’, a company active in the manufacture of components for automotive airbag modules.170 The assessment of customer foreclosure concerns conducted by the Commission in this case is discussed in some detail below.
4.4.1 Case Study: WLR/BST In WLR/BST, the Commission established that there was no horizontal overlap between the merging parties. However, BST produced a particular product, flat airbag fabric, which was an input to a product produced by SCI, cut and sewn airbag cushions (CSCs). Its competitive assessment therefore addressed whether this vertical relationship could give rise to customer foreclosure, a concern that was raised by a number of competitors of BST in the supply of flat airbag fabric.172 The Commission found that in the upstream market for flat airbag fabrics, BST was the EEA-wide market leader with a market share in 2005 of 45–50%.173 On the downstream market for CSCs, there were two main ‘independent’ suppliers, SCI and Aerazur, with market shares in ‘independent’ sales of 40–45% and 35–40% respectively. However, in addition to these ‘independent’ suppliers, there were producers of CSCs that were vertically integrated in the production of airbag modules and that supplied CSCs exclusively to their own in-house operations.174 The Commission noted that ‘independent’ CSC producers were subject to strong competitive pressure from airbag module suppliers’ own in-house production of CSCs, and that on the basis of total CSC production the share of SCI was 15–20%.175 The Commission considered a theory of harm in which the merged entity would foreclose rival flat airbag fabric suppliers’ access to a substantial CSC customer base, and thereby reduce the ability of rival fabric producers to compete. The Commission noted a number of factors that undermined the merged entity’s ability to engage in customer foreclosure. The Commission found that it was difficult for a CSC producer to switch fabric 171
170 171 172
173
See Case COMP/M.4389 – WLR/BST, Commission decision of 5 December 2006. Case COMP/M.4389 – WLR/BST, Commission decision of 5 December 2006. The Commission also considered and dismissed possible competitive concerns arising from input foreclosure, although it did not receive any complaints from third parties with concerns related to input foreclosure. See ibid., para. 30. 174 175 Ibid., para. 24. See ibid., para. 10. See ibid., para. 23.
vertical mergers: customer foreclosure
451
supplier quickly. This difficulty arose from ongoing production program commitments on both sides and the need for certain certification procedures to be followed if switching was to occur.176 The Commission nevertheless considered the consequences of a decision by the merged entity’s downstream division to source flat fabric exclusively from its upstream division. The Commission found that only one competitor of BST currently depended on SCI for more than 10% of its flat fabric turnover.177 It also found that most of BST’s main competitors were large diversified companies that also produced a range of technical fabrics for applications other than airbag modules, and that could be expected to withstand a sudden drop in flat fabric sales without undergoing severe financial difficulties. The Commission noted that if SCI sourced exclusively from BST, this would bring only about 10–20% of the total EEA flat fabric production in-house. The Commission recognised that given that BST’s capacity was fully utilised, and could not be increased economically in the short term, a consequence of this demand being switched to BST would be that it would have to reduce sales to SCI’s competitors. This displaced supply would leave unsatisfied demand for fabric from SCI’s competitors, which could only be met by BST’s competitors. Moreover, the Commission found that, given that SCI’s purchases of flat airbag fabrics represented only 15–25% of total EEA demand for flat airbag fabrics, BST’s competitors would still have access to a substantial customer base and noted in this context that overall EEA demand for airbag products was growing at around 5% per annum.178 The Commission also considered the role played by customers of CSCs, airbag module manufacturers. It found that airbag module manufacturers exercised considerable influence on the choice of fabric supplier even when outsourcing CSC production, and that they had a preference for multiple sourcing of fabric in order to maintain sufficient competitive alternative suppliers.179 Given that the Commission did not find the merged entity to have the ability to foreclose, it did not need to carry out a detailed analysis of the merged entity’s incentives in WLR/BST. However, a particular concern raised by one of BST’s fabric competitors was that the merged entity, which controlled its major cost input (i.e. fabric), would opt to cut prices of CSC in order to expand its share of the 176 177
Case COMP/M.4389 – WLR/BST, Commission decision of 5 December 2006, para. 33. 178 179 See ibid., para. 33. See ibid., para. 35. See ibid., para. 36.
452
non-horizontal mergers
CSC market; and that this would create a ‘backlash’ demand for cheaper fabric from other CSC’s producers. This would allegedly squeeze BST’s fabric competitor’s margins, perhaps causing some to exit the market.180 Interestingly, rather than responding to such concerns by noting that to the extent the merger gave rise to downward pressure on both CSC and fabric prices, this should be welcomed as a benefit of the merger, the Commission dismissed the concern by stating that the merged entity would not have an incentive to adopt such a strategy. It countered this concern by noting that if airbag module producers saw that SCI was enjoying a lower fabric transfer price from BST post-merger, they would themselves request a corresponding price cut from BST for their fabric prices for in-house production. The Commission then reasoned that since fabric margins are much higher than CSC margins, the overall trade-off of a low-price CSC strategy would be negative for the merged entity. This is because any profits from increased volumes of CSC sales resulting from lower prices for the 10– 30% of BST output that goes to SCI would have been outweighed by lower margins on the 70–90% of BST output sold to third parties. It should be noted that this analysis is somewhat unconventional. It relies on BST having to lower its prices to rivals to its downstream division as a result of its downstream division becoming more aggressive in competition with those rivals. The Commission did not carry out a more conventional analysis of the incentives of the merged entity to engage in customer foreclosure that would have needed to take account of the costs of reducing purchases from certain upstream suppliers, which, absent the foreclosure strategy, would have been purchases the downstream division preferred to undertake.
5.
Conglomerate mergers
The term ‘conglomerate merger’ refers to any merger that is neither horizontal (i.e. between actual or potential competitors) nor vertical (i.e. between a supplier and a customer). Conglomerate mergers therefore include mergers between suppliers of complementary products and mergers in which the parties supply products that are neither complementary nor substitutable.181 In most circumstances, such mergers will 180 181
Case ibid., para. 37. In other words, an increase or decrease in the price of one product does not change the demand for the other product.
conglomerate mergers
453
not raise competition concerns.182 However, it is possible that under certain circumstances a merger between producers of complementary products or related products that are purchased by the same set of customers could give rise to anti-competitive effects. The principal theory of harm applicable to conglomerate mergers is that the merger would allow for market power in one activity to be used as a means of reducing competition in a related market. This is often referred to as ‘leveraging’ of market power. It could also be the case, however, that a merger raises the possibility that the merged entity could pursue conduct that allows it to protect existing market power that, absent the merger, might have become subject to more effective competition. In both cases, the concern is that an increase in product range allows the merged entity to take some action that lessens the competitive constraints that it faces, thereby giving rise to a unilateral incentive to increase prices to the detriment of consumers. Generally, this action will involve the conditioning of sales in a market currently served by one merging party on sales in another market currently served by the other merging party via tying and/or bundling.183 The term ‘pure bundling’ is used to describe a situation in which products are sold together in fixed proportions and are not available independently of one another. For example, a pure bundle of good A and good B implies that a customer can buy good A and good B together, but cannot buy either good A or good B on its own. ‘Mixed bundling’, on the other hand, describes a more general situation in which goods are sold together in fixed proportions but are, in addition, available independently of one another. Foreclosure concerns may arise where a bundle is sold for a price lower than the sum of the prices of the individual goods in question. Pure bundling can be considered as a special case of mixed bundling whereby the prices of the individual goods in the package are so high that no customer would purchase the products separately. The term ‘tying’ is used to describe a situation in which customers that purchase a particular good (the tying good) are also required to purchase another good (the tied good). For example, tying of good B to good A implies that if a customer purchases good A, they must purchase good B rather than good C, which might otherwise have been an alternative to good B. Unlike the case of a pure bundle, the consumer need not 182 183
See EC Non-horizontal Merger Guidelines, para. 92. See ibid., para. 93.
454
non-horizontal mergers
purchase good A and good B in fixed proportions but may choose the volume of B they purchase.184 Tying can be achieved through designing the tying good so that it does not work when used with products competing with the tied good (technical tying) or through contracting such that a customer purchasing the tying good must agree to purchase the tied good and not products competing with the tied good (commercial tying). The general concern in relation to each of these practices is that the merged entity will reduce rivals’ profits to such an extent that they are no longer able to exert an effective competitive constraint on the merged entity, with the result that consumers are harmed. As with vertical mergers, the EC Non-horizontal Merger Guidelines put forward a three-stage approach for the assessment of whether a conglomerate merger could be considered to give rise to anti-competitive foreclosure.185 These stages relate to the questions of whether a merged entity would have the ability to foreclose its rivals, whether it would have the incentive to do so and whether foreclosure would have a detrimental effect on competition that gives rise to consumer harm. In what follows we consider in turn the factors relevant to these three analytical elements. As with the discussion of vertical mergers, it should be noted that these stages are closely interrelated and in practice a single factor, such as the likely responses by rivals or the existence of economies of scale, may be relevant to the consideration of each of these questions. Having considered the factors relevant to the Commission’s threestage assessment, we then provide a hypothetical example that explains in more detail how a merged entity’s ability and incentive to foreclose, and the likelihood that foreclosure leads to consumer harm, depends on these factors. Finally, we present case studies of the Intel/McAfee,186 GE/Amersham187 and Google/DoubleClick188 mergers and comment on a number of additional cases in which the Commission considered conglomerate concerns.
184
185 186 187 188
Under some definitions of tying, a customer purchasing the tying good (product A) may be permitted to choose a zero volume of the tied good (product B), but is nonetheless prevented from purchasing alternatives to the tied good (product C). See EC Non-horizontal Merger Guidelines, para. 94. Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011. Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004. Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008.
conglomerate mergers
455
5.1 Ability to foreclose in conglomerate mergers The first step in the assessment of whether the merged entity has the ability to foreclose through any action (such as tying or bundling) is consideration of whether it is actually feasible in practical terms for the merged entity to undertake the action in question. In many cases, an examination of the features of a market reveals that a particular strategy, such as pure bundling, would not be a practical option for the merged entity. It may be that two goods have different procurement procedures, different supply chains or different procurement timelines. For example, one good may represent a one-off purchase that is undertaken only once in a decade, whereas the other good may be purchased regularly over the course of the lifetime of the first good. This might make it much more difficult to engage in any form of commercial tying of the second good to the first good. In particular, it may be the case that the products are purchased by different customers, even if at first sight this is not obvious given that the products in question are components that end up in the same system. This may be because different companies make decisions over which products to purchase for different elements of a single system. For example, it may be that an aircraft manufacturer chooses one component for the aircraft but an airline chooses the engine at a later stage via separate competitions that take place each time an airline places an order for the aircraft. Selling to different customers often makes it impossible to engage in tying or bundling.189 Alternatively, there may be different buying groups within a single company that are responsible for buying the different products. Whilst this may not make tying or bundling impossible, it could in practice make it significantly more difficult since it may require the seller to effect changes to the procurement procedures currently in place. In many cases the technical tying of one product to another will not be an option. In some cases this will be obvious (e.g. in the case of a merger between a producer of tea bags and tea cups) but in other cases involving products of a more technical nature this will be a more complex question. Moreover, rather than eliminating compatibility with rival products outright, the merged entity might be able to merely diminish the compatibility of its products with that of its rivals such that their functionality 189
It may still be possible to engage in technical tying with separate buyers through the implementation of compatibility standards.
456
non-horizontal mergers
is degraded when used with rival products instead of with other products produced by the merged entity.190 One factor that may in certain cases influence the assessment of whether bundling or tying of one product to another is feasible is whether there is evidence that these practices have been carried out in the past by companies whose range includes the products in question. There is an important distinction between the question of whether the merged entity has the ability to engage in bundling and tying and the question of whether the merged entity has the ability to foreclose. A firm may face no material impediments to linking sales of one of their products to sales of another (e.g. offering a discount on one product in their product line to any customer that purchases another of its products). Importantly, however, such a firm will not necessarily have the ability to foreclose. Specifically, it will not necessarily be able to materially reduce the revenues or increase the costs of rivals in a way that undermines their competitiveness. For that to be the case a number of further conditions must hold.191 First, the merged entity must have a significant degree of market power in at least one of the markets in which it is active. Market power could arise as a result of the merged entity offering a highly differentiated product for which rival products are not considered a good substitute or as a result of rivals facing significant capacity constraints. Absent market power, it is not clear how any action undertaken by the merged entity could materially affect the extent to which its rivals can offer effective competition. If the firm is subject to effective competition for each of its products then it will be able to induce customers to choose a bundle of its products only via lower prices, which is the essence of competition and not a cause for concern. Only where a firm attempts to use pre-existing market power in one area to harm rivals and thereby lessen competition in another activity can anti-competitive foreclosure potentially arise. It is interesting to note that bundling and tying may be observed in markets in which firms do not have a significant degree of market power and in cases where the profitability of these strategies clearly does not derive from foreclosure of rivals. This leads to the important observation, discussed in relation to the example below, that a merger could give rise 190
191
This could be considered as a case of mixed bundling, whereby quality is increased, rather than price lowered, when the merged entity’s products are purchased together rather than separately. Each of these is discussed further in the example in Section 5.4 below.
conglomerate mergers
457
to the ability and incentive to bundle products in cases where this strategy is profitable even absent any foreclosure of rivals. Second, there must be a large common pool of customers for the products that are the subject of the leverage theory.192 If the products in question are consumed alone by many customers, then the merged entity will be less able to affect the demand of individual products produced by its rivals through any strategy that links the sales of its products to one another. In addition to these conditions, a number of other factors are relevant for the questions of whether the merged entity has the ability to foreclose. First, it is relevant to consider whether, by reducing the demand for its rivals’ products, the merged entity can materially affect their costs (or alternatively, the quality of their products) such that rivals are less able to offer effective competition or are even forced to exit the market entirely. A particular action may reduce a rival’s sales significantly. However, if there are no significant economies of scale, for example, this reduction in demand may have no impact on the rival’s incentive to remain in the market and to try to win incremental business at a similar price to that which it would have offered absent the action in question. Given that foreclosure, as defined in the EC Non-horizontal Merger Guidelines, incorporates the notion of rivals’ ‘ability and/or incentive to compete’,193 it is important to take into account rivals’ cost structures at the stage of assessing the merged entity’s ability to engage in foreclosure as well as at the stage of assessing its incentive to foreclose and the effect of foreclosure.194 Second, it is relevant to consider whether there are counterstrategies available to rivals that might be expected to undermine the merged entity’s ability to foreclose. For example, it may be possible for rivals whose product ranges mirror that of the merged entity to replicate the bundling or tying strategy of the merged entity. If none of the merged entity’s rivals presently produce all the goods that form part of the merged entity’s tie or bundle, then it may be necessary for them to merge with one another or enter ‘teaming arrangements’, whereby they jointly supply a particular customer that demands both elements of the bundle. By offering a similar bundle, the merged entity’s rivals may be able to neutralise any competitive advantage that might otherwise have arisen. 192 193
EC Non-horizontal Merger Guidelines, para. 100. 194 Ibid., para. 18. Ibid., para. 101.
458
non-horizontal mergers
5.2 Incentive to foreclose in conglomerate mergers Even if the merged entity has the ability to foreclose through a strategy of bundling or tying, it will not necessarily find it profitable to do so. This is because tying and bundling can give rise to immediate costs to the merged entity that may not outweigh their benefits in terms of generating greater sales volumes or allowing higher prices in the market into which it is attempting to leverage market power.
5.2.1 Costs of bundling and tying The costs of bundling and tying will depend on the extent to which the customer base for each of the products in question is interested in also purchasing the other product. In the extreme scenario when there are no common customers for the products in question (as would be the case, for example, with respect to GE’s jet engines and its medical diagnostic products), pure bundling will likely give rise to significant losses as many customers that would otherwise have chosen to purchase one of the products in the pure bundle prefer to forgo consumption of that product rather than purchase the pure bundle. With regard to mixed bundling, the absence of common customers will imply that demand for this mixed bundle will be low. Moreover, it will eliminate the incentive to charge higher prices for the products when purchased in isolation because doing so will only cause customers to switch to rival stand-alone products rather than to the mixed bundle. Even in the case where there are common customers for the products in question (as may be the case, for example, with respect to two products within GE’s medical diagnostic product range), bundling can entail significant losses. A pure bundle may cause certain customers that previously purchased one of the products from the merged entity and one of the products from a rival to purchase both products from the merged entity’s rivals rather than purchase the pure bundle. In addition, if the merged entity can only successfully foreclose its rivals by offering a very low price for the pure or mixed bundle, then it may be that the profits sacrificed to implement the foreclosure strategy are substantial. 5.2.2 Benefits of bundling and tying Tying and bundling may lead to certain benefits to the merged entity. Importantly, whilst the benefits of tying and bundling may depend on the effect of foreclosure (discussed further below), in principle there may be instances where tying or bundling that leads to the foreclosure of rivals
conglomerate mergers
459
would have been profitable absent any foreclosure effect. In this case, to the extent that the bundling gives rise to foreclosure that subsequently allows the merged entity to increase price, this will constitute an additional benefit to the merged entity arising from the bundling strategy. Mixed bundling in particular, may be profitable for the merged entity even absent foreclosure if there is a significant degree of complementarity between the products of the merging parties or if mixed bundling increases the scope for the merged entity to engage in price discrimination. If the products of the merging parties are complementary then, as discussed in Section 2.1, the merged entity will take account of the fact that a reduction in the price of one good will boost demand for the other good and vice versa (the so-called Cournot effect). The extent to which a given price reduction will be profitable will depend both on the extent to which it boosts demand for the complementary product and the margin earned on the incremental sales of the product whose demand is boosted by the price reduction. The greater the increase in demand for the other good (that is, the higher degree of complementarity) and the greater the margin earned on incremental sales of the other good, the greater will be the incentive to cut the price of the first good.195 In order to ensure that any increase in demand for complementary goods is captured by the merged entity and not by competitors, the merged entity may have an incentive to make the reduction in the price conditional on the purchase of the complementary products from the merged entity (i.e. offer the complementary products in a mixed bundle). Even if the products sold by the merged entity are not complementary, an incentive to engage in mixed bundling to a pool of common customers may arise due to the increased scope for price discrimination it can offer. It has been shown that offering a mixed bundle is in theory profitable for a monopolist of multiple products under surprisingly general demand conditions, if the firm can otherwise only charge a single price for each product.196 Indeed, it can be shown that for any two goods, A and B, that have independent distributions of values (i.e. a consumer
195
196
This is analogous to the case of a merger between substitutes, in which the incentive to increase the price of one product is affected by both the increase in demand for the other product and the margin earned on incremental sales of the other product. See R. P. McAFee, J. McMillan and M. Whinston, ‘Multiproduct Monopoly, Commodity Bundling, and the Correlation of Values’ (1989) 103 Quarterly Journal of Economics 371–83.
460
non-horizontal mergers
with a high willingness to pay for A relative to other consumers will not necessarily have a high willingness to pay for B relative to other consumers) a monopolist can obtain higher profits by offering a mixed bundle of A and B at a discount to the sum of the optimal monopoly prices of the two products.197 To see why this is the case, suppose the monopolist sets the price of A and B optimally, i.e. at the level for which marginal revenue equals marginal cost. At these price levels a marginal reduction in the price of either good has no impact on profits (the reduction in margin on sales to existing customers is just offset by the incremental sales made). Now suppose the monopolist offers a marginal price reduction for good B that is conditional on the purchase of A. This means the firm earns a lower margin on sales to consumers that would have bought B anyway, but it also gives rise to an increase in demand for B. These two effects just offset one another. This is because the price of B was previously set optimally and, importantly, those customers that obtain the discount (i.e. only those that purchase A) are representative of the overall demand for B – they are no more or less likely to have a high willingness to pay for B than other consumers due to the independence of the demand for A and B. Crucially, however, the conditional discount will also cause an increase in demand for A. This is because certain customers that did not buy A previously, will do so now that it enables them to purchase B at a marginally lower price. Since we noted that the effect on profits was neutral before taking into account the increased sales of A, these extra sales of A must increase profits over and above the profit levels obtainable absent the mixed bundle.198 Importantly, however, if firms are already engaging in price discrimination pre-merger, both the Cournot effect and any benefit related to further prospects for price discrimination disappear. In the case that each customer is charged its willingness to pay for each unit produced by the merging parties, the merged entity will not be able to boost sales by decreasing the price it charges to any individual customer implying that no Cournot effect can arise post-merger. Moreover, to the extent
197
198
The result is surprisingly general because the condition that two goods have independent distributions of valuations is likely to be satisfied for the vast majority of pairs of unrelated goods chosen at random, such as apples and pens. For a further explanation of this result and an interesting discussion of why we do not in practice see a plethora of mixed bundles containing unrelated goods, see B. Nalebuff, ‘Bundling, Tying, and Portfolio Effects’ (DTI Economics Paper No. 1, 2003).
conglomerate mergers
461
that each merging party is already able to price discriminate effectively through its individual price negotiations with customers, mixed bundling will offer no further scope for price discrimination. The discussion above illustrates that tying and bundling can be profitable absent a foreclosure motive, and indeed may lead to price reductions that benefit consumers. Often, however, the benefits of tying and bundling against which the costs of these strategies must be compared may depend to a significant extent on the effect of foreclosure. If tying and bundling is very effective at causing foreclosure, and foreclosure gives rise to an opportunity for the merged entity to significantly increase its prices and/ or sales volumes, then the merged entity may have an incentive to foreclose even if it is not profitable in the short run. Whether the foreclosure strategy is profitable over the long term will then depend on whether the benefits from foreclosure outweigh the costs. The factors that influence the effect of foreclosure are discussed further below in Section 5.3. As a final remark, if bundling involves a sacrifice of profit to drive out competitors, then it becomes similar in character to predatory pricing. As discussed in more detail in the example below at Section 5.4, any theory of harm involving short-term profit sacrifice must specify why such a strategy would become attractive or feasible post-merger given that we did not previously observe the merging parties engaging in predatory pricing pre-merger. Thus, while it could be alleged that a merged firm could offer a reduction in the price of good A to induce customers to choose good B over rivals’ substitutes for B, the supplier of B could, premerger, offer an equivalent price cut on B to win sales from rivals. In this regard, it may be relevant to consider whether the merger makes the strategy significantly less costly (due to the demand boosting effect caused by the complementarity of the products, which is not taken into account by either party pre-merger) and, therefore, whether tying and bundling enables the merged entity to commit to more aggressive pricing behaviour in a particular market than it could have done absent the ability to link sales in that market to sales in another market.
5.3 Effects of foreclosure in conglomerate mergers Bundling and tying by a merged entity may result in significant harm to its competitors by reducing demand for those rivals’ products. However, it is important to recognise that all competitive conduct is intended to win share from rivals, and so anti-competitive foreclosure must be
462
non-horizontal mergers
distinguished from ordinary competition on the merits. Anti-competitive foreclosure is said to arise only when harm to competitors directly gives rise to harm to consumers. The merged entity’s rivals may experience a significant reduction in revenues without having their ability or incentive to compete with the merged entity for incremental business materially affected. Moreover, even if the competition exerted by the merged entity’s rivals is adversely affected, consumers may gain overall as a result of lower prices arising from efficiencies generated by the merger. As with theories of harm based on customer foreclosure, a link must be established between rivals’ reduced demand and a reduced ability to compete. Nevertheless, in certain circumstances foreclosure may allow the merged entity to increase or protect market power in one of its markets via either exclusion or marginalisation of rivals. Increases in market power may come about if rivals are no longer viable as a result of the foreclosure strategy and are forced to exit. Similarly, the protection of market power may result from entry that would have otherwise occurred being deterred. More generally, increases in or the protection of market power may result from important rivals to the merged entity being marginalised. Absent countervailing factors such as buyer power or new entry (or indeed re-entry by previously marginalised rivals) that disciplines the merged entity’s pricing decisions over the longer term, consumers may be harmed overall by higher prices than would have otherwise prevailed. The marginalisation of rivals could occur if variable costs fall significantly with the level of output produced. In that case a strategy that reduces the output of rivals may increase their variable costs. This in turn could make them unwilling to offer low prices in competition with the merged entity for incremental customers. Alternatively, if rivals expect to earn lower revenues in the future as a result of the merged entity’s foreclosure strategy, they may be deterred from investing in R&D that they might otherwise have found attractive. This may imply that in future their products will be less attractive to customers than they would have been absent foreclosure, and that as such they exert a less significant constraint on the merged entity’s prices. It is often difficult in practice to assess how foreclosure will affect the incentives of rivals to invest in R&D. It will sometimes be relatively clear, however, that a reduction in revenues in a particular product or geographic market is unlikely to affect the incentives of a rival that is active across multiple product or geographic markets to invest in R&D.
conglomerate mergers
463
Even if certain competitors are marginalised (or even exit the market) as a result of foreclosure, this will not necessarily have a material impact on the competitive constraints remaining on the merged entity. It is only if a set of rivals that together represent an important competitive constraint on the merged entity are marginalised that foreclosure could give rise to concerns regarding a lessening of effective competition. It may be that firms offering single products in each of the markets served by the merged entity continue to offer effective competition even if certain rivals are marginalised, particularly if the rivals that are marginalised have higher costs or offer less attractive products. As noted above, rivals may have the ability and incentive to respond to any attempt by the merged entity to engage in foreclosure by employing counterstrategies (or even counter-mergers) that undermine the ability of the merged entity to foreclose. Such strategies are also relevant to consider at the stage of assessing the effect of foreclosure. In certain cases sophisticated customers who have a clear interest in ensuring that they are served by competitive markets may play an important role in mitigating the anti-competitive effects of foreclosure. In addition to strategies such as threatening to sponsor entry, customers may exert buyer power through effectively sponsoring actions on the part of the merged entity’s rivals, such as encouraging them to offer competing bundles through teaming arrangements, which serve the long-term interests of the buyer in terms of maintaining competition. The role that may be played by large customers who can strategically condition their purchasing decisions on their long-term interests is overlooked in much of the economic literature concerning bundling and tying. This factor, however, may be a very important consideration in many conglomerate merger cases involving industrial products. Importantly, even if rivals are deemed unable to propose a competitive response that undermines the merged entity’s foreclosure strategy and are forced to exit as a result, competition is only harmed if the exit enables the merged entity to implement sustainable increases in its prices. In certain cases, such price increases will be rendered unattractive by the prospect of inviting re-entry by rivals that previously exited due to foreclosure. For example, a rival that is induced to exit from a particular geographic market due to foreclosure, but remains active in other geographic markets, may find re-entry relatively easy and profitable if the merged entity attempts to exploit market power by charging high prices. Finally, the assessment of efficiencies may form an important part of the assessment of the overall effect of a conglomerate merger. Efficiencies
464
non-horizontal mergers
might arise from economies of scope whereby the merged entity can reduce their costs by selling a wider range of products. Alternatively, customers may gain significantly from being able to purchase multiple products from a single supplier. Moreover, the internalisation of any pricing externality due to the bringing together of sales of complementary goods may represent an efficiency that directly benefits consumers. The source of competitive concern (e.g. low priced mixed bundles that give rise to foreclosure of rivals) may be closely related to efficiencies that lead to consumer benefits in the short term. If an investigation suggests that consumers will generally gain in the short term as a result of lower prices, then only in exceptional circumstances should the merger be prohibited on the basis that they might be harmed in the long run due to the marginalisation of rivals. Although these circumstances might in principle arise if the benefits to consumers in the short term were shown to be small and the long-term harm was shown to be large and relatively certain, such circumstances are likely to be rare in practice.
5.4 Foreclosure in conglomerate mergers: an example In this section we consider a highly stylised example that highlights a number of features relevant to the assessment of the ability and incentive to engage in foreclosure through bundling.199 Suppose that consumers buy systems that are made up of two types of component, A and B. A consumer derives no value from either component A or component B on its own but derives some value from consuming them together. Component A is produced by two competing Firms, A1 and A2, whilst component B is produced independently by two competing Firms, B1 and B2. In this example, we consider the implications of a merger between A1 and B1. Pre-merger, consumers can choose to purchase one of four systems as follows: A1B1; A1B2; A2B1; or A2B2, where A1B1 denotes a system that comprises an A component from Firm A1 and a B component from Firm B1. Suppose that the systems are differentiated so that at equal prices for each system as a whole, some consumers will prefer one system whilst others will prefer another.
199
This example is in the spirit of Choi (2008) (below). While we make use of some key intuitive results from Choi, we do not provide a detailed explanation of the assumptions behind these results here. See J. P. Choi, ‘Mergers with Bundling in Complementary Markets’ (2008) 3 Journal of Industrial Economics 553–77.
conglomerate mergers
465
Suppose that each of the four components is produced at zero marginal cost. However, suppose that each component producer has to pay a one-off fixed cost of 8,000 every year to stay in the market.200 Pre-merger, each component is priced at 50 and the demand for each of the four systems is 100 units (implying that total demand for each component is 200 units). The total revenues of each component producer is 10,000 (since each sells 200 units at a price of 50), which is sufficient to justify the fixed cost of staying in the market of 8,000. Post-merger, the choice set of the consumer depends on whether the merged entity decides to engage in some form of bundling. If the merged entity decides not to engage in any form of bundling, all four system choices will remain available post-merger. If the merged entity decides to engage in mixed bundling, then all four system choices will continue to be available but the price of A1 will be lower when it is bought together with B1 than when it is bought on its own to be assembled by the consumer with a B component produced by the merged entity’s rival, B2. Similarly, the price of B1 will be lower when it is bought together with A1 than when it is bought on its own to be assembled by the consumer with an A component produced by the merged entity’s rival, A2. If the merged entity decides to engage in pure bundling, then it will now not be possible for a consumer to purchase A1 and assemble it with B2, or for a consumer to purchase B1 and assemble it with A2. Therefore there will only be two systems available for consumers to choose: A1B1 or A2B2. We consider the merged entity’s ability and incentive to pursue strategies of mixed and pure bundling in turn.
5.4.1 Mixed bundling Let us assume that the merged entity offers a mixed bundle of A1 and B1, such that whereas the total price of the system A1B1 was previously 100 (since each component was priced at 50), this system is now priced at 80. Prior to the merger, each producer does not take into account the fact that a reduction in price of its component will boost demand for complementary components. The merger allows this effect to be taken into consideration (i.e. the pricing externality is internalised), allowing the 200
This could relate to the costs of maintaining the plants in which each component is produced.
466
non-horizontal mergers
price of the mixed bundle to be lower than the combined price of the individual components if they were priced independently. However, the merged entity decides to increase the prices of the standalone products to 60. It takes account of the fact that if it increases the prices of the stand-alone products, some of the lost demand for systems A1B2 and A2B1, which are now more expensive due to the higher prices of both A1 and B1, will be recaptured by its low priced system A1B1. In response to this, suppose that producers of the rival components, A2 and B2, decide to reduce the price of their components from 50 to 45. This price reduction comes about for two reasons. First, the lower price for the mixed bundle implies that the stand-alone providers have an incentive to reduce their prices so that each of the systems that contain one of their own components remains competitive with the merged entity’s mixed bundle. Second, the higher price for the individual components A1 and B1 implies that each of the stand-alone producers has an incentive to reduce price in order to reduce switching from the ‘mix and match’ systems, A2B1 and A1B2, to the mixed bundle offered by the merged entity. For example, there is an incentive to reduce the price of A2 to minimise the loss of switching from the system A2B2, since the substitutable product A1B1 is cheaper, and a further incentive to reduce the price of A2 to minimise the loss of switching from the system A2B1, which, absent a reduction in the price of A2, will seem particularly unattractive given the higher price of the stand-alone component B1. However, they do not reduce their prices to as low as 40 (i.e. the implicit price of each component within the merged entity’s mixed bundle) as each stand-alone firm does not take account of the fact that a small reduction in the price of its component boosts demand for the component of the other firm.201 As a result of these pricing decisions, each of the ‘mix and match’ systems, A2B1 and A1B2, now has a total price of 105. This is because they include an individual component from the merged entity, priced at 60, and an individual component from a stand-alone producer, priced at 45. 201
These pricing movements are directionally the same as those found in Choi (2008). Specifically, Choi’s model shows (given the assumptions regarding demand and cost conditions) that mixed bundling following the merger would have the following implications for prices: first, the price of the bundle post-merger is lower than the sum of the pre-merger component prices; second, the merged firm’s prices for individual components are higher with mixed bundling; and third, the independent firms also cut their prices. See J. P. Choi, ‘Mergers with Bundling in Complementary Markets’ (2008) 3 Journal of Industrial Economics 553–77.
conglomerate mergers
467
Table 5.3 System prices pre-merger and post-merger under mixed bundling System
Pre-merger price
Post-merger price
A1B1 A2B2 A1B2 A2B1
100 100 100 100
80 90 105 105
The system comprising components from each of the stand-alone producers, A2B2, has a total price of 90, because each component has a price of 45. The system comprising components from the merged entity, A1B1, has a total price of 80, the price of the mixed bundle. Table 5.3 summarises the total prices of each of the four systems premerger and in the post-merger scenario in which mixed bundling is adopted. We now consider how the volume demanded for each system might change from the pre-merger situation in which demand for each system was 100. Since the systems are differentiated, some consumers will prefer a more expensive system to a cheaper one, and, indeed, some still prefer a ‘mix and match’ system. Let us assume that demand for the mixed bundle increases significantly to 170 units, as certain consumers switch from each of the other systems to the cheapest system available. The price reduction in the A2B2 system causes it to retain some customers that might otherwise have switched to the mixed bundle and also to win some sales from the higher priced ‘mix and match’ systems, and as a result experiences a more modest increase in demand to 110 units. The demands for the mix and match systems, A2B1 and A1B2, fall to 60 units as many consumers switch to either A1B1 or A2B2. Table 5.4 summarises the total demand for each of the four systems pre-merger and in the postmerger scenario in which mixed bundling is adopted.202 It is worth noting at this stage that the short-term welfare implications of the above vary across the heterogeneous consumers. It is clear that those who preferred system A1B1 at pre-merger price levels benefit since they can now buy the same system at a lower price. Similarly, those who preferred system A2B2 are better off since this system has also reduced in 202
Directionally, these changes in demand are similar to those found in equilibrium in Choi: see J. P. Choi, ‘Mergers with Bundling in Complementary Markets’ (2008) 3 Journal of Industrial Economics 553–77.
468
non-horizontal mergers
Table 5.4 System demand pre-merger and post-merger under mixed bundling System
Pre-merger demand
Post-merger demand
A1B1 A2B2 A1B2 A2B1
100 100 100 100
170 110 60 60
price. However, those that purchased a mix and match bundle, and continue to do so despite the new higher price, are worse off. The overall welfare effect is therefore ambiguous.203 It is now possible to determine the revenues of each firm. The revenue earned by the merged firm derives from three sources. First, revenues from sales of system A1B1 are equal to 13,600 (170 units at a price of 80 each). Second, the revenues from sales of A1 components as part of the demand for the system A1B2 are equal to 3,600 (60 units at a price of 60 each). Third, and symmetrically, the revenues from sales of B1 components as part of the demand for the system A2B1 are equal to 3,600 (60 units at a price of 60 each). The total revenue of the merged entity is therefore equal to 20,800. The total revenue of each stand-alone component provider can also be calculated. Revenues for the producer of A2 components come from two sources. First, revenues from the sales of A2 components as part of the demand for the system A2B2 are 4,950 (110 units at a price of 45 each). Secondly, revenues from the sales of A2 components as part of the demand for the system A2B1 are 2,700 (60 units at a price of 45 each). The total revenues of the producer of A2 components is therefore 7,650 (4,950 + 2,700). Since the producer of B2 components is in a symmetrical position, its revenues are also 7,650. The revenues of the stand-alone providers have therefore fallen from pre-merger levels of 10,000, to 7,650 as a result of the merger and mixed bundling strategy. Recall, that we assumed that there was a fixed cost of 8,000 of maintaining the plants for producing each component. On this basis, the producers of the stand-alone components would prefer to exit 203
Although a key result of Choi (ibid.) is that the overall consumer welfare effect can be negative if substitutability of different systems is sufficiently high. The price of the standalone products is raised to a greater extent in this case: see ibid.
conglomerate mergers
469
the market as a result of the merger and mixed bundling strategy rather than pay the fixed cost of 8,000. If the producers of the stand-alone components exit, then the merged entity will face a reduced competitive constraint for each of its components and will find it profitable to increase prices, thus harming all consumers. In this stylised example, the presence of the fixed cost implies that mixed bundling gives rise to the exit of the merged entity’s rivals, and a lessening of competition in the longer term, despite the possible benefits to consumers in the short term from lower prices (recall that both systems A1B1 and A2B2 fell in price). Note that, in this case, the profits of the merged entity in the short term (i.e. prior to any exit) are higher than the pre-merger profits of A1 and B1 combined. Pre-merger, producers of A1 and B1 each had revenues of 10,000. Post-merger, the merged entity earns revenues of 20,800 (implying that each division can earn a revenue share of 10,400). The firms have an incentive to merge and engage in mixed bundling even in the absence of any motive to cause the exit of rivals (e.g. they would have this incentive even if the fixed cost were only 6,000 rather than 8,000, implying that rivals would remain in the market). The fact that the mixed bundling will lower rivals’ revenues to such an extent that they exit is a ‘bonus’ in this example, which may enable the merged entity to increase profits to an even greater extent over the longer term. Some authors refer to a situation in which bundling would be profitable in the short term even absent a foreclosure motive as non-strategic bundling and to one in which bundling is only profitable given foreclosure as strategic bundling.204 In practice, however, the distinction between strategic and nonstrategic bundling is not clear-cut. The ability and incentive to foreclose stem from a number of features of the example.205 First, there is a common pool of customers for A-components and Bcomponents. In fact, every consumer of an A-component is also a consumer of a B-component. If the components were consumed alone by many consumers (e.g. certain consumers demand an A-component 204
205
See, e.g., B. H. Kobayashi, ‘Does Economics Provide a Reliable Guide to Regulating Commodity Bundling by Firms? A survey of the economic literature’ (2005) 1(4) Journal of Competition Law & Economics 707–46. As noted above, it is difficult to isolate factors that give rise to an ability to foreclose from factors that give rise to an incentive to foreclose in this example, because a factor that affects the extent to which there exists an ability to foreclose may simultaneously affect the extent to which there is an incentive to foreclose.
470
non-horizontal mergers
but have no interest in consuming a B-component) then there would be a reduced incentive to increase the individual component prices for A1 and B1, since these face competition from individual components A2 and B2, and the stand-alone providers would have a lower incentive to reduce their prices. Moreover, with a smaller common pool of customers for the A-component and the B-component, the ability of the merged entity to engage in foreclosure will be reduced since it will be less able to affect the demand for the individual components A2 and B2 through mixed bundling. Certain consumers that purchase A2 but no B-component will not be persuaded to switch to A1 if they have no need for B1, because the stand-alone A2 component is cheaper than the A1B1 system as a whole. Second, we have implicitly assumed that the merged entity has some degree of market power. When increasing the prices of stand-alone components A1 and B1, it need only worry about the diversion of demand that this might cause to the single system that contains none of the merged entity’s components, A2B2. If however, the merged entity faces competition from a number of producers of A-components and B-components it may be less inclined to increase the price of its individual components since this may cause diversion to a variety of different systems. For example, with a third pair of components, A3 and B3, rather than there being one system containing none of the merged entity’s components, A2B2, there are now four, A2B2, A2B3, A3B2 and A3B3. To the extent that A3 is a close substitute for A1, a customer that previously purchased system A1B2 may switch to system A3B2 rather than system A1B1 following an increase in the price of the stand-alone component A1 (even given the reduction in price of system A1B1). In addition, the existence of many producers of A-components may imply that B2 has a reduced incentive to decrease its price in response to an increase in the price of the stand-alone A1 component. One way to see why this might be the case is to consider that there is less need to reduce price to remain competitive since whereas before B2 was only part of one system that had not become more expensive, A2B2, it is now part of two such systems, A2B2, and A3B2. Therefore, when the merged entity faces more competition for its products, it will generally have a lower ability to reduce the revenues of rivals through any mixed bundling strategy. Third, there is an extreme form of economy of scale driven by the existence of the fixed cost. If rivals cannot cover the fixed cost they are induced to exit. More generally, the foreclosure effects of mixed bundling could still be a concern if there were a less extreme form of scale economies. For example, if firms faced incremental costs that reduced
conglomerate mergers
471
as output increased then a reduction in rivals’ demand due to mixed bundling could give rise to an increase in their incremental costs and thus a reduction in the extent to which they constrain the prices of the merged entity. It could also be that the mixed bundling strategy reduces rivals’ incentives to invest in cost-reducing R&D.206 However, in instances where the demand for rivals’ products and their profits today do not affect their ability and incentive to compete in the future it is not clear why mixed bundling should be viewed to give rise to marginalisation of rivals and less effective competition in future. Specifically, in our stylised example, with a lower fixed cost (e.g. of 6,000), there would be no exit and therefore no loss of competition in the market (subject to the caveat that some consumers may be worse off in our example due to the higher prices of systems A1B2 and A2B1). The existence and nature of scale economies is a highly relevant factor in the analysis of firms’ ability and incentive to foreclose, and the ultimate effect of foreclosure. Fourth, we have assumed that the systems required by end-users comprise only two components, both of which are produced by the merged entity. Suppose that instead of consisting of an A-component and a B-component, a system also has a C-component, which is not produced by either of the merging parties. Suppose that the C-component represents the vast majority of the total price of a system, with A- and B-components representing only a small fraction of the total price. This will, in effect, limit the extent to which A-components and B-components are complementary. It would take a very large reduction in the price of an A-component to materially reduce the price of the system, and thereby increase the demand for systems and, consequently, increase the demand for extra B-components. This will therefore reduce the extent to which there is a Cournot effect in which there is a significant incentive to sell the bundle of A- and B-components at a low price. Benefits from cutting price will exist in so far as this gives rise to stealing of market share from rival suppliers of A- and B-components but not in terms of increasing the overall market size (which is the only benefit originally identified by Cournot, who considered a merger between monopoly providers). Fifth, we have assumed that pre-merger the producers of A1 and B1 had to set a single price to all customers. Suppose instead that component prices are negotiated on a customer-by-customer basis. In the extreme 206
See J. P. Choi, ‘Mergers with Bundling in Complementary Markets’ (2008) 3 Journal of Industrial Economics 553–77.
472
non-horizontal mergers
case where firms negotiate prices with perfect information and are able to negotiate the division of any surplus, the Cournot effect discussed above disappears.207 Suppose that a particular customer is willing to pay up to 60 for component A1 and 55 for component B1. If each component producer can perfectly price discriminate, then they will charge prices of 60 and 55 respectively for components A1 and B1 to that particular customer. If the two firms merge, they would set the price of the system A1B1 equal to 115 (the sum of the customer’s willingness to pay for each component). There would be no incentive to reduce the price to this customer given that the customer is willing to pay this price.208 Sixth, we have not considered the possibility of counterstrategies on the part of the stand-alone component providers. For example, they could invest in product improvements or cost reductions to make systems containing their components more competitive with the mixed bundle offered by the merged entity. Another possible counterstrategy available to producers of A2 and B2 is that they engage in a merger themselves and offer a competing mixed bundle. This would restore symmetry in terms of the demand for each system. However, it is important to consider whether a counter-merger would be profitable. It could be that a counter-merger by the stand-alone providers is not immediately profitable because internalisation of the Cournot effect creates an intense form of competition between the rival systems that reduces profits for the producers of A2 and B2 relative to the situation in which they remain separate.209 Nonetheless, whilst the counter-merger might reduce the short-term profits of A2 and B2 it would also reduce the profits of the merged entity producing A1 and B1. To the extent that over the longer term this would mean that they are less likely to fall behind in R&D, the rival stand-alone providers may prefer to merge.210 Moreover, 207
208
209
210
Perfect information is an extreme assumption. However, this case clearly shows the effect of relaxing the assumption of ‘one price for all’ when products are sold individually. The point that individual negotiation with customers can make the benefits of bundling disappear was raised by economists advising GE in the GE/Honeywell merger (see, e.g., B. J. Nalebuff, ‘Bundling and the GE–Honeywell Merger’ (Yale SOM Working Paper No. ES-22, September 2002) and, as discussed further in Section 5.5.3 below, was recognised by the Commission in Google/DoubleClick (See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008, para. 353). See J. Church, ‘Conglomerate Mergers’, in Issues in Competition Law and Policy (American Bar Association Section of Antitrust Law, 2008), vol. II, pp. 1503–22. See B. J. Nalebuff, ‘Bundling and the GE–Honeywell Merger’ (Yale SOM Working Paper No. ES-22, September 2002).
conglomerate mergers
473
even if producers of components A2 and B2 did not want to offer a competing bundle, powerful customers, who would stand to gain if they did, may try to force them to do so.211 Counterstrategies, such as a merger by stand-alone providers, could seriously undermine the ability of the merged entity to foreclose, the incentive to foreclose and any adverse effect of mixed bundling on end consumers. Seventh, it is possible that a profitable opportunity would exist for a company to purchase the mixed bundle and sell the components unbundled at a profit. For example, it could purchase the mixed bundle at a price of 80 and sell each component at a price of 50, effectively neutralising the mixed bundling strategy. Unless the merged firm can ensure that this type of arbitrage does not take place, then it will find it impossible to engage in mixed bundling.212
5.4.2 Pure bundling As noted above, if the merged entity decides to engage in pure bundling, then it will now not be possible for a consumer to purchase A1 and assemble it with B2, or for a consumer to purchase B1 and assemble it with A2. Therefore, there will be only two systems available for consumers to choose: A1B1 or A2B2. In particular, suppose that the merged entity can commit to the pure bundling strategy by making its products available only as an integrated system, with the components technically incompatible with those offered by other suppliers.213 Note that at pre-bundling prices, the aggregate demand for all four components must fall. This is because the effect of bundling is to make the systems A1B2 and A2B1 unavailable and certain consumers that purchased these systems will prefer to purchase no system rather than either of the two that remain available. The merged entity has an incentive to lower the price of the pure bundle for two reasons. First, it has fewer infra-marginal units due to the loss of demand for A1 components that formed part of the demand for the system A1B2 and the loss of demand for B1 components that formed 211 212
213
See ibid. This is similar to the condition that for a firm to engage in price discrimination, e.g. by selling to some consumers at a high price and to some consumers at a low price, it must be able to stop the consumers that purchase at a low price from re-selling to the high price consumers at a price that undercuts the firm. The importance of being able to commit to a strategy of pure bundling is discussed in Whinston (1990), and is also discussed further below. See M. D. Whinston, ‘Tying Foreclosure, and Exclusion’ (1990) 80 American Economic Review 837–59.
474
non-horizontal mergers
Table 5.5 System prices pre-merger and post-merger under pure bundling System
Pre-merger price
Post-merger price
A1B1 A2B2
100 100
90 110
part of the system A2B1. This means that a price reduction is less costly since it will lower margins on a smaller number of units than previously. Second, it has an incentive to reduce price due to the Cournot effect given that A1 and B1 are complementary products. Suppose that absent any reaction by rivals the merged entity chooses to set a price for the pure bundle of 80. However, the effect of pure bundling may cause the stand-alone providers to increase their prices.214 Previously, any decrease in the price of A2 would have boosted its demand as consumers purchased more A2B1 systems and A2B2 systems. However, under pure bundling any price decrease will give rise to a smaller increase in demand since consumers cannot purchase A2 as part of the A2B1 system any more. Put differently, the upside of competing aggressively (in terms of additional sales that can be made) has become less pronounced due to the presence of the pure bundle. Given that it faces less elastic demand, the producer of A2 components has, therefore, an incentive to increase its price. A similar situation applies to the producer of B2. Suppose that both producers of A2 and B2 components choose to increase their price from 50 to 55. Given the price increase by rivals, the merged entity may be expected to choose further to increase its prices in response to the lessened competitive constraint that it faces. We suppose that given the responses of its rivals the merged entity settles on a price for the pure bundle of 90. Table 5.5 summarises the total prices of each of the A1B1 and A2B2 systems pre-merger and in the post-merger scenario in which pure bundling is adopted. We now consider how the volume demanded for each system changes from the pre-merger situation in which demand for each of the four systems was 100. Given that A2B1 systems and A1B2 systems are no 214
See J. Church, ‘Conglomerate Mergers’, in Issues in Competition Law and Policy (American Bar Association Section of Antitrust Law, 2008), vol. II, pp. 1503–22.
conglomerate mergers
475
Table 5.6 System demand pre-merger and post-merger under pure bundling System
Pre-merger demand
Post-merger demand
A1B1 A2B2
100 100
180 110
longer available, some previous consumers of these systems will switch to the pure bundle offered by the merged entity and some (smaller number) will switch to the alternative system available, A2B2. Importantly, however, some consumers will no longer choose to purchase any of the systems at the prices on offer. The total demand for components will fall as a result of the number of systems available reducing from four to two. Suppose the demand for the pure bundle increases from 100 to 180 and the demand for the alternative system offered by the stand-alone providers increases from 100 to 110. Table 5.6 summarises the total demands for each of the two systems pre-merger and in the post-merger scenario in which pure bundling is adopted. The revenues earned by the merged entity now derive solely from the sale of the pure bundle. It sells 180 systems at a price of 90 each and therefore earns revenues of 16,200. The total revenues of each stand-alone component provider derive solely from the sale of their components as part of the demand for one system, A2B2. Each stand-alone component provider sells 110 components at a price of 55 each and therefore earns revenues of 6,050. The revenues of the stand-alone providers have therefore fallen from pre-merger levels of 10,000 to 6,050. This is in large part due to the fact that many consumers simply leave the market given the inability to mix and match A2 with B1 and A1 with B2. As in the case of mixed bundling, the producers of the stand-alone components would prefer to exit rather than pay the fixed cost of 8,000. If the producers of the stand-alone components exit, then the merged entity will face a reduced competitive constraint for each of its components and will find it profitable to increase prices, thus harming all consumers. Note that in this case, the profits of the merged entity in the short term (i.e. prior to any exit) are lower than the pre-merger profits of A1 and B1 combined. Pre-merger, producers of A1 and B1 each had revenues of
476
non-horizontal mergers
10,000. Post-merger, the merged entity earns revenues of 16,200 (implying that each division can earn a revenue share of 8,100). The attractiveness of the pure bundling strategy in this particular example therefore relies on the exit of rivals, which enables the merged entity to increase profits over the long term. The merged entity would not engage in pure bundling absent this foreclosure motive. If bundling involves a sacrifice of short-term profit to drive out competitors, it becomes similar in character to predatory pricing. If we do not observe predatory pricing with respect to the merging parties’ products pre-merger, any theory of harm involving short-term profit sacrifice must explain why this sacrifice becomes profitable post-merger. The argument that the merged entity will be able to ‘cross-subsidise’ the losses on one product with profits gained as a result of market power on another product is not satisfactory in this regard. If forcing the exit of rivals through predatory conduct were profitable in the long run (noting, of course, that this could be illegal under Article 102),215 then it is not clear why it would only be undertaken if there were cross-subsidisation from another product. To the extent that the strategy represents an attractive use of profits earned on the other good, so it would also have been an attractive use of funds (retained profits or borrowing) premerger. Note, however, that there may indeed be a reason why the merger creates an opportunity to force the exit of rivals that did not exist premerger. Suppose that the producers of A2 and B2 components are just about to decide whether to incur the fixed cost of staying in the market. Suppose further that if the producers of A1 and B1 components could make a binding commitment to charging ‘low’ prices next period, the producers of A2 and B2 components would choose to exit rather than incur the fixed cost. However, the producers of A1 and B1 components 215
The CFI criticised the Commission for its failure to take due account of the deterrent effect of Article 102 TFEU on the merged entity’s incentives to engage in bundling postmerger in the GE/Honeywell judgment: ‘Given the extreme nature, from a commercial perspective, of the behavior, it was incumbent on the Commission to take into account the effect which the Community-law prohibition on abuses of a dominant position might have had on the merged entity’s incentive to implement such practices. Since the Commission failed to do that, it made an error of law.’ See Case T-210/01, General Electric v. Commission [2005] ECR II-5575, para. 425. More generally, the ‘Article 102 defence’ may not be effective. The Commission has subsequently held that not all foreclosure strategies can be easily or swiftly detected and that in any event intervention can take place only after the negative effects have materialised: see Case COMP/M.3440 – ENI/EDP/GDP, Commission decision of 9 December 2004.
conglomerate mergers
477
cannot make such a commitment. If the producers of A2 and B2 components choose to stay in the market by incurring the fixed cost, it will then be more profitable for the producers of A1 and B1 to renege on any threat to set ‘low’ prices and accommodate their rivals’ decisions to stay in the market through higher prices. Pure bundling can, however, remedy this commitment problem under certain circumstances. Suppose that whilst the producers of A1 and B1 cannot commit to low prices pre-merger (given that price decisions are easily reversible), the merged entity can take some irreversible action that commits it to offering the pure bundle. In particular, suppose it makes A1 and B1 components technically compatible only with one another and not with the complementary components offered by the merged entity’s rivals. Suppose further that it would be very costly to reverse this decision. The irreversibility of this decision means that the rivals will now exit rather than pay the fixed cost because they know that unlike the threat of low prices, the merged entity cannot undo the decision to pure bundle even if to do so would be profitable in the event that the rivals did in fact stay in the market. Ex ante, the decision to pure bundle is not short-run profit maximising. However, the short-run profit-maximising decisions of the merged entity given that it has committed to offer the pure bundle will result in revenues for the rivals that are insufficient to cover the fixed costs of staying in the market. This induces rivals to exit and therefore leads to potentially higher profits for the merged entity over the long term. A forward-looking merged entity may find the decision to offer the pure bundle profitable overall.216 It should be noted that the attractiveness of the pure bundling strategy need not rely on the exit of rivals, as it did in the example above if systems are relatively homogeneous. If systems are relatively homogeneous the loss of demand as a result of consumers leaving the market will be smaller (i.e. there will be more switching to the bundles that remain available) and so pure bundling could in fact be profitable in the short term, to the extent that it causes rivals to increase their prices, even absent the exit of rivals.217 216
217
Whinston was the first to point out that bundling can be used as a device to enable a monopolist to commit to more aggressive behaviour against an entrant and that this commitment can deter entry: see M. D. Whinston, ‘Tying Foreclosure, and Exclusion’ (1990) 80 American Economic Review 837–59. If systems are relatively homogeneous the loss of demand is less severe from pure bundling as consumers will substitute to the pure systems rather than leave the market altogether. Moreover, under mixed bundling independent firms lower their
478
non-horizontal mergers
As with mixed bundling the ability and incentive to foreclose stem from certain features of this example. First, there is a common pool of customers for A-components and Bcomponents. If the components were consumed alone by many consumers (e.g. certain consumers demand an A-component but have no interest in consuming a B-component) then the pure bundling strategy could cause a substantial loss of demand for the merged entity as a result of the elimination of the stand-alone product. Moreover, the effect of pure bundling may be to drive consumers to the stand-alone products produced by rivals, undermining the ability of the merged entity to foreclose rivals through this strategy. Second, the exit of rivals is contingent on the fixed cost of remaining in the market. As noted above in the discussion of mixed bundling, with a lower fixed cost (e.g. of 6,000), there would be no exit and therefore no loss of competition in the market. The merged entity would no longer have the ability to marginalise its rivals. Moreover, since pure bundling is only profitable in this example in so far as it causes rivals to exit (and competition to subsequently be reduced), the merged entity will no longer have the incentive to pure bundle given the lower fixed cost. As in the case of mixed bundling, rivals’ scale economies will be a relevant factor in the consideration of the merged entity’s ability and incentive to foreclose and in the ultimate effect of foreclosure. Third, we have ignored practical considerations that may undermine the ability of the merged entity to engage in pure bundling. As noted above, these practical considerations include the possibility that different components have different procurement procedures, different supply chains or different procurement timelines. In addition, there may be technical barriers to making the components incompatible with those offered by rivals (such as sophisticated customers that can adapt or modify products to restore compatibility).
component prices but under pure bundling they raise them. Pure bundling therefore leads to higher prices when systems are homogeneous. This also means that the profits of the independent firms are greater under pure bundling relative to mixed bundling when the systems are relatively undifferentiated. In any event, even where systems are more differentiated so that the short-term profits from mixed bundling are higher than those from pure bundling, the merged firm may still prefer to engage in pure bundling because the negative impact on independent firms’ profits is greater under pure bundling than mixed bundling when systems are relatively differentiated. J. Church, ‘Conglomerate Mergers’, in Issues in Competition Law and Policy (American Bar Association Section of Antitrust Law 2008), vol. II, pp. 1503–22.
conglomerate mergers
479
Fourth, suppose that, pre-merger, component prices are negotiated on a customer-by-customer basis. This can have significant implications for the extent to which pure bundling increases the ability and incentive of the merging parties to foreclose competitors relative to the pre-merger situation. Suppose that a particular customer prefers component A1 to component A2 and will choose A1 over component A2 as long as the price differential between the two components is not greater than 5. On the other hand, suppose that the customer prefers component B2 to component B1 and will choose B2 over component B1 as long as the price differential between the two components is not greater than 5. Premerger, suppose that the outcome of negotiations are such that the customer chooses A1 and B2 because neither the producer of A2 nor that of B1 is willing to undercut the prices of its rivals by 5 (which would enable it to win the customer’s business) but rather it is only willing to undercut by marginally less than 5. Post-merger, the merged entity considers trying to switch the customer’s demand from B2 to B1 by pure bundling A1 and B1. Note, however, that if the merged entity is to induce the customer to purchase the pure bundle, it will need to price the pure bundle at a level below the sum of the pre-merger prices it offered for the individual components. Given that pre-merger the producer of B1 had no incentive to reduce its price a small amount to win this business, it is not clear why the merged entity should have an incentive to lower the implicit price of B1 in order to make the customer accept the bundle. As put by Barry Nalebuff:218 If customers would prefer some products from A and others from B, then the combined firm will continue to offer the individual goods at their premerged prices. Forcing a bundle on the consumer can only lower firm A’s profits. In effect, it would have to subsidize its disadvantage using profits it could have earned from products where it is strongest. This is no different from selling individual components at a loss – a strategy it can do but would choose not to, even without bundling.
Fifth, as discussed above in relation to mixed bundling, counterstrategies may be available to the merged entity’s rivals that undermine the exclusionary effects of pure bundling.
218
Barry Nalebuff is Milton Steinbach Professor of Management at the Yale School of Management and was an economic adviser to GE during the Commission’s investigation in the GE/Honeywell merger. See B. J. Nalebuff, ‘Bundling and the GE–Honeywell Merger’ (Yale SOM Working Paper No. ES-22, September 2002).
480
non-horizontal mergers
5.5 Assessment of foreclosure in conglomerate mergers: EU case law There has been only one case in the relevant period 2004–2010 where the Commission identified conglomerate concerns leading to remedies: Intel/ McAfee.219 This is discussed in Section 5.5.1. Nonetheless, there are a number of cases in which the Commission has considered conglomerate theories of harm. Two of these, GE/Amersham220 and Google/DoubleClick,221 are discussed in detail in Sections 5.5.2 and 5.5.3 below, respectively. We also make a number of references (without providing detailed discussion) to some additional cases in which the Commission has considered conglomerate concerns, including Johnson&Johnson/Guidant,222 IBM/Telelogic,223 Thomson Corporation/Reuters Group,224 Omya/Huber PCC225 and Siemens/VATech.226 As noted above, although conglomerate overlaps encompass all markets that are not horizontally or vertically related, the Commission’s practice shows that the focus is on those mergers that bring together economic complements or products that are sold to the same set of customers. As with vertical foreclosure concerns, significant market power pre-merger is a necessary condition for a finding of harm. Before considering the Commission’s analysis in recent cases it is worth commenting briefly on the GE/Honeywell merger.227 A great deal had been written on the economic issues in this case and as such we do not discuss this case in detail here.228 The case was an important one because it represented a major divergence in policy with respect to non-horizontal 219 220 221 222
223 224
225 226 227 228
Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011. Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004. Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008. Case COMP/M.3687 – Johnson&Johnson/Guidant, Commission decision of 25 August 2005. Case COMP/M.4747 – IBM/Telelogic, Commission decision of 5 March 2008. Case COMP/M. 4726 – Thomson Corporation/Reuters Group, Commission decision of 19 February 2008. Case COMP/M.3796 – Omya/Huber PCC, Commission decision of 19 July 2006. Case COMP/M.3653 – Siemens/VATech, Commission decision of 13 July 2005. Case COMP/M.2220 – GE/Honeywell, Commission decision of 3 July 2001. The following represents a sample of the commentaries on the economics of this case: B. J. Nalebuff, ‘Bundling and the GE–Honeywell Merger’ (Yale SOM Working Paper No. ES-22, September 2002); D. S. Evans, and M. Salinger, ‘Competition Thinking at the European Commission: Lessons from the aborted GE–Honeywell merger,’ George Mason Law Review (May 2002), 489–532; J. P. Choi, ‘The Economics and Politics of International Merger Enforcement: A case study of GE/Honeywell Merger’, in V. Ghosal and J. Stennek (eds.), Political Economy of Antitrust (Emerald Group Publishing, 10 Apr 2007); X. Vives and G. Staffiero, ‘The GE–Honeywell Merger in the EU’, in B. Lyons
conglomerate mergers
481
mergers between the US Department of Justice, which cleared the deal, and the European Commission, which prohibited it. Moreover, on appeal of the Commission’s decision to the CFI,229 the Commission was heavily criticised by the court for its analysis of the conglomerate effects, which was found to be ‘vitiated by manifest errors’.230 During the Commission’s investigation economists on both sides argued at length over the short-term incentive to engage in mixed bundling (particularly on account of the Cournot effect, see Section 5.4 above). However, the Commission concluded that this issue could be left open since the merged entity would in any event have an incentive to engage in bundling in order to marginalise its rivals. The Commission did not deem it necessary to rely on any specific economic model or analysis to support its bundling theories, explaining that it had: evaluated the theoretical premises of mixed bundling as presented to it in the economic analyses submitted by the parties and third parties. The various economic analyses have been subject to theoretical controversy, in particular as far as the economic model of mixed bundling, prepared by one of the third parties, is concerned. However, the Commission does not consider reliance on one or the other model necessary for the conclusion that the packaged deals that the merged entity will be in a position to offer will foreclose competitors from the engines and avionics/non-avionics markets.231
The Commission did not specify how the incentive to engage in foreclosure would arise post-merger when prices are set through individual negotiation and an incentive to marginalise rivals by predatory pricing had not existed pre-merger (see the Nalebuff critique noted in Section 5.4 above). The Commission emphasised that ‘[t]he ability of the merged entity to cross-subsidise its various complementary activities and to engage in profitable forms of packaged sales will have an adverse effect on the profitability of competing producers of avionics and non-avionics products, as a result of market share erosion.’232 Subsequent to the CFI’s critique of its analysis in GE/Honeywell, and the publishing of the EC Non-horizontal Merger Guidelines, the
229 230
231 232
(ed.), Cases in European Competition Policy, The Economic Analysis (Cambridge University Press, 2008). Case T-210/01, General Electric v. Commission [2005] ECR II-5575. The CFI upheld the prohibition decision only in so far as it agreed with parts of the analysis concerning horizontal aspects of the merger. Case COMP/M.2220 – GE/Honeywell, Commission decision of 3 July 2001, para. 352. Ibid., para. 398.
482
non-horizontal mergers
Commission has adopted a more evidence-based approach to conglomerate issues. Where the Commission has considered the merged entity’s incentive to engage in foreclosure in cases since 2004, it has recognised that tying and bundling may result in significant costs. The Commission has also placed emphasis on the question of whether rivals would continue to pose an effective competitive constraint even if the merged entity engaged in tying or bundling.
5.5.1 Intel/McAfee In January 2011, the Commission conditionally cleared the merger between Intel, the leading producer of central processing units (CPUs) and chipsets, and McAfee, a provider of IT security solutions. The Commission required remedies to allay its concerns regarding nonhorizontal relationships between the parties’ products, and specifically, its concern that the merger would give rise to anti-competitive foreclosure of rival providers of security solutions. At the outset, the Commission noted that there were important linkages between computer hardware (CPUs and chipsets) and security solutions. First, there was a technological link. The Commission noted that security software interacts to a greater extent than other software directly with the hardware level. This implied that security software vendors (‘SSVs’) needed access to up-to-date, accurate and complete interface information on new CPUs and chipsets in order to be able to develop new security software.233 Second, the Commission noted the commercial link between the parties’ products. In particular, it noted that every device running using Intel hardware would in principle need some form of security software and that the same intermediaries, for example, Original Equipment Manufacturers (‘OEMs’) would be involved in the decisions over which CPUs to use and which security to install.234
233
234
Interface information was also required to optimise the software with regard to performance and power consumption since the running of security processes may significantly increase the workload on the CPU and affect the available performance of the computer. In addition, the Commission noted that certain features of IT security could be more effectively enabled in hardware than in software and that the partial embedding of security solutions in hardware may lead to more robust and/or faster security solutions. See Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011, paras. 18 and 19. See ibid., para. 21.
conglomerate mergers
483
As a result of these linkages, the Commission considered whether Intel would have the ability and incentive to bundle or tie Intel CPU chipsets and McAfee security solutions; and/or to degrade the interoperability between Intel’s hardware and security solutions on the one hand and the products of competing providers of security solutions on the other. The Commission found that its conclusion from its 2009 Intel antitrust decision regarding Intel’s dominance continued to hold.235 This conclusion was based on persistently high market shares, which for CPUs were higher than [70–80]% in each of the desktop, notebook and netbook sub segments and which for chipsets were higher than [70–80]% in a general chipset market and higher than [90–100%] in the possible ‘aftermarket’ of chipsets for Intel-compatible CPUs. In addition, the Commission considered that these markets were subject to very high barriers to entry and expansion.236 The Commission noted that Intel faced one competitor in CPUs, AMD, with a share of [10–20%] and another, Via, whose position was insignificant in terms of volume shipped.237 In particular, the Commission referred to its 2009 Intel antitrust decision, which had found that a potential entrant faces significant intellectual property barriers and has to engage in substantial initial research and development and production investment to be able to start production of x86 CPUs.238 The Commission noted that McAfee was one of the few security technology companies active in practically all areas of the security technology spectrum, serving end-consumers, SMEs, and large private corporations and governmental organisations.239 In particular, it had a share of [10–20%] in end point security, in which it was the second largest worldwide player behind Symantec [30–40%]. The Commission noted that barriers to expansion in the enterprise segment were significant due to the need for brand recognition, a range of products, services and support, and trust that the company will continue to innovate.240 In the consumer segment, the Commission found a clear division between the players with access to the OEM channel (Symantec, McAfee and Trend Micro) and the other players without such access.241 The Commission also noted that the three largest players benefited from network 235 236
237 240
COMP/C-3 /37.990 – Intel, (‘the 2009 Intel antitrust decision’). See Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011, para. 62. 238 239 See ibid., para. 72. See ibid., para. 91. See ibid., para. 63. 241 See ibid., para. 102. See ibid., para. 104.
484
non-horizontal mergers
effects since the more users an SSV has, the higher its chances to detect new malware, in particular when those users are distributed evenly across different geographic areas.242 Complainants to the merger raised three forms of foreclosure concerns, namely degradation of interoperability between Intel’s hardware and security solutions on the one hand and the products of competitors on the other, technical bundling/tying (‘technical tying’) and commercial bundling strategies. The degradation of interoperability and commercial bundling concerns are discussed in turn below,243 followed by a brief summary of the remedies implemented to allay these concerns. 5.5.1.1 Degradation of interoperability The Commission found that Intel would have the ability to degrade the interoperability of its hardware with security solutions provided by McAfee’s rivals.244 It found that developers need a range of information and a number of tools from the CPU manufacturers to develop and optimise their software for a given CPU.245 The Commission focused on ‘positive discrimination’, whereby McAfee would be given preferential access to the procedures of the 242 243
244
245
See ibid., para. 106. Refusal to provide interoperability information to third parties that enables rivals to link their products to the merged entity’s products is sometimes considered as a special case of technical tying, although in Intel/McAfee the Commission considered these concerns separately. The Commission’s findings with respect to technical tying (integrating McAfee technology into Intel’s hardware) were similar to those in the discussion of degraded interoperability and we therefore do not discuss these separately. As with degradation of interoperability, the Commission concluded that Intel would have the ability and incentive to engage in technical tying and that this practice would likely result in harm to consumers. The Commission’s discussion of technical tying can be found at Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011, paras. 175–221. The Commission has considered interoperability issues in a number of cases. In IBM/ Telelogic the Commission examined whether the merged entity would have the ability to refuse to provide interoperability information and found that it would have such ability. Indeed, it considered that ‘any software vendor whose product’s output must somehow become the input for other software products in principle has this ability’ although it noted that the ability to withhold interoperability information applied to only new software projects and not existing ones. See Case COMP/M.4747 – IBM/Telelogic, Commission decision of 5 March 2008, para. 255. Other cases in which interoperability has been considered include Case COMP/M.3083 – GE/Instrumentarium, Commission decision of 2 September 2003, and Case COMP/M.2861 – Siemens/Drägerwerk/JV, Commission decision of 30 April 2003. See Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011, para. 140.
conglomerate mergers
485
CPU.246 It considered that this form of discrimination would be difficult to detect and that it would not be possible for McAfee’s rivals to react to discrimination by ‘reverse engineering’ Intel’s hardware components.247 Regarding the incentive to degrade interoperability, the complainants pointed to a specific statement by Intel CEO Paul Otellini, which was interpreted as Intel’s explicit intention to favour McAfee by granting it better access to Intel features or technologies than its competitors.248 The Commission also highlighted internal documents relating to the integration of McAfee products.249 Critically, the Commission found that Intel could implement the hypothesised strategy without losing CPU revenues. Specifically, it noted that the only way CPU revenues would decrease would be if end-users show such a loyalty to an SSV other than McAfee that they would switch their demand to an alternative CPU supplier, which the Commission found unlikely.250 Finally, the Commission found regarding the effect of foreclosure that the strategy would result in the exclusion of McAfee’s competitors from Intel’s platform, as a result of security software from other vendors being perceived as less effective. 251 The Commission noted that innovation incentives of rivals would be impacted with a likely reduction in consumer choice. The Commission also suggested that a degradation of McAfee’s interoperability with other chipsets or CPUs could further protect Intel in the CPU and chipset markets by increasing barriers to entry.252 5.5.1.2 Commercial bundling The Commission considered whether the merged entity would have the ability and incentive to engage in pure commercial bundling (whereby CPUs and security software are sold exclusively together) and mixed commercial bundling (whereby the security software would be offered at a discount when customers buy both products from Intel/McAfee). Intel raised a number of practical difficulties that it would face in implementing a bundling strategy, including that CPUs for consumer PCs are typically negotiated on a 246 248
249 252
247 See ibid., para. 142. See ibid., para. 145. ‘While we’ll still work with the Symantecs and Microsofts and Nortons of the world, we’re also going to make sure that the best possible solution is Intel on Intel or Intel on McAfee in this case and that it’s architected to run best together.’ See Case COMP/ M.5984 – Intel/McAfee, Commission decision of 26 January 2011, para. 154. 250 251 See ibid., para. 161. See ibid., para. 159. See ibid., para. 166. See ibid., para. 172.
486
non-horizontal mergers
quarterly basis while security software contracts typically cover multiple years and that the pricing terms for security software are complex.253 However, while the Commission noted that contract cycles and durations for CPUs and security solutions are currently different and that negotiations are currently also often led by different teams for each of these products, it found that given Intel’s power over OEMs it could change negotiation and contract formats.254 The Commission found likely that Intel has the ability to enter into a foreclosure strategy through a commercial bundling between its chipsets/CPUs and McAfee’s endpoint security.255 Regarding the incentive to engage in the hypothesised strategies, the Commission placed some emphasis on economic analysis submitted by Intel. First, this analysis claimed that the share of sales of all endpoint security software which could be shifted to McAfee by a tying or bundling strategy is low (less than 1%). The analysis took account of the relatively low proportion of PCs that both ship with an Intel processor and ship with preloaded security software trials provided by McAfee competitors (5–10%), took account of the fact that only 0–5% of these trials actually result in a security software subscription, and that some software sales are due to subscription renewals rather than new subscriptions.256 Second, the analysis considered the critical loss accrued by Intel which would render commercial bundling unprofitable.257 The analysis compared Intel’s average profit on CPUs and chipsets that would be lost on each computer for which the OEM switches away from Intel, with the McAfee’s expected lifetime profit on an incremental trial, which would be won on every PC with an incremental trial. The analysis found that bundling would be unprofitable if [less than 1%] of Intel’s hardware sales were switched away from Intel.258 The Commission adopted a similar framework as Intel and maintained a number of the analysis’ assumptions. It found that for a reasonable range of such assumptions, the critical loss stemming was higher than the critical loss submitted by Intel, but remained limited nonetheless. It therefore concluded that the incentives of Intel to engage in a commercial 253
254 257
258
See Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011, para. 229. 255 256 See ibid., para. 237. See ibid., para. 238. See ibid., para. 268. The critical loss was defined as the share of lost hardware sales that would result in a loss of profit to the merged firm that exactly offsets the profit gained by it from incremental security software trials. See ibid., para. 269. See ibid., para. 270.
conglomerate mergers
487
bundling might indeed be weakened by the risk of losing CPU sales to OEMs.259 The Commission concluded that Intel would have limited incentives to engage in such a strategy if pursued in isolation.260 Finally, regarding the effect of foreclosure, the Commission found that a commercial bundling strategy would be unlikely to give rise to anticompetitive foreclosure of rivals. In contrast to the other strategies that the Commission considered, commercial bundling could only affect sales to consumers through Original Equipment Manufacturers (OEMs).261 The Commission noted, however, that rivals would be able to reach customers in other ways, for example by focusing on direct sales or on the enterprise segment. 5.5.1.3 Remedies In order to address the Commission’s concerns regarding the degradation of interoperability and technical tying, Intel offered a set of behavioural remedies. The Commission accepted that these were best suited to remedy its concerns in the present case and set out the conditions that the remedy would need to comply with in order to be effective. Regarding interoperability, Intel’s main commitment was to disclose instructions and interoperability information on a royalty free basis.262 Regarding tying, Intel would be able to provide bundled solutions, but OEMs would be able to effectively disable any bundled Intel security functions without this degrading the quality or performance of rival security solutions.263 The commitments would be valid for a period of 5 years, would be monitored by a trustee and would be accompanied by a dispute settlement mechanism including a fast track arbitration procedure.
5.5.2 GE/Amersham The Commission had to consider a number of issues related to the ability and incentive to engage in tying and bundling during its investigation of the GE/Amersham264 merger. The products produced by the parties that were sufficiently closely related to trigger potential conglomerate concerns were GE’s diagnostic imaging (DI) equipment, used to generate images of the human body for 259 262
263 264
260 261 See ibid., para. 276. See ibid., para. 289. See ibid., para. 285. See Case COMP/M.5984 – Intel/McAfee, Commission decision of 26 January 2011, paras. 289 and 337–40. See ibid., paras. 341 and 342. Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004.
488
non-horizontal mergers
medical diagnosis, and Amersham’s diagnostic pharmaceuticals (DPs), used to enable or enhance the clarity of the image produced by DI equipment. The products were related in the sense that they had common customers (hospitals) and formed complementary components in systems used by hospitals for diagnosis. The Commission defined a number of different relevant product markets for DI equipment that related to the five main DI modalities of imaging equipment (X-ray, Computed tomography (CT), Magnetic resonance (MR), Ultrasound (U/S) and Nuclear imaging (NI)), but left open whether these could be further subdivided.265 The Commission left open the exact scope of the DP markets although it considered that the parties’ proposed delineation, between contrast agents and radioactive pharmaceuticals, would be too broad. This is because it considered that there was limited demand-side and supply-side substitutability between contrast agents used with certain different DI modalities and, moreover, for certain procedures involving NI, organ specific DPs were required that could each constitute a separate market.266 Regarding the geographic market, the Commission left open whether markets could be considered national or EEA-wide but considered the possibility that markets should be defined narrowly in its competitive assessment by considering the parties’ positions in various national markets.267 Whilst the Commission noted that GE had a high market share in certain product areas and in certain national markets, it noted that the market investigation did not attribute to GE any specific characteristics that would provide it with an important economic advantage over competitors.268 The Commission noted that Amersham held very high market shares in certain DPs but also that, given the importance of R&D in the markets concerned, high market shares might only be temporary due to the potential for leapfrogging by competitors introducing new products.269 The Commission considered whether the merged entity would have the ability and incentive to foreclose competition by leveraging any premerger market power from one market to another. It considered two strategies in particular. First, it considered mixed (commercial) bundling of DI equipment and DPs, whereby the merged entity would offer discounts on these products when purchased together.270 Second, it 265 268 270
266 267 Ibid., para. 9. Ibid., paras. 13–14. Ibid., para. 18. 269 Ibid., para. 23. Ibid., para. 28. The Commission also considered pure bundling (or what it termed ‘forced bundling’) of DI equipment and DPs, whereby the merged entity would no longer make its products
conglomerate mergers
489
considered technical tying (whereby its products would work optimally or exclusively with one another). The Commission focused on those procedures where there was significant complementarity between DI equipment and DPs. 5.5.2.1 Mixed bundling The Commission considered a theory of harm in which the merged entity would offer a bundle of DI equipment and DPs at a lower price than the sum of the individual components when purchased alone with the aim of excluding competitors. Third parties alleged that, as a result, they would face reduced revenue streams and a reduced ability and incentive to invest in R&D, which might gradually lead to their marginalisation and exit.271 The parties pointed to a number of factors that raised questions about the feasibility of such a strategy. They noted that there were significant differences in the procurement procedures and supply chains for DI equipment and DPs and that the products also had different procurement timelines.272 DI equipment represents a one-off major capital expenditure that is made infrequently given its minimum lifecycle of 10 years, while DPs are purchased on an ongoing basis throughout the lifecycle of the DI equipment. The Commission noted that, whilst these considerations did not necessarily preclude commercial bundling, they would complicate its successful implementation.273 The Commission noted that, whilst the parties had high market shares in some product markets in certain member states, it considered that these did not in themselves indicate dominance in the context of these
271
272
273
available on a stand-alone basis but only as a bundle. It was quick to dismiss this possibility, however, given that the merged entity would lack an incentive to engage in such a practice since it would deny it significant sales of DPs to current users of non-GE equipment and significant sales of DI equipment to users that would prefer to continue using non-Amersham DPs. See ibid., para. 43. Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004, para. 34. In a number of cases, the Commission has determined that a particular bundling strategy would not be feasible because the customers purchasing the products in question are not the same. For example, in Johnson&Johnson/Guidant, the Commission noted with respect to a bundling strategy involving different areas, such as endovascular, interventional cardiology, and Cardiac management system devices (defibrillators and pacemakers), that ‘a broader bundling involving products of different areas is hardly feasible as customers are generally not the same’. See Case COMP/M.3687 – Johnson&Johnson/Guidant, Commission decision of 25 August 2005, para. 342. Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004, para. 35.
490
non-horizontal mergers
markets and highlighted features such as the bidding nature of the markets, the lumpy and infrequent procurement of equipment, existence of credible alternative suppliers, leapfrogging through innovation, absence of switching costs and countervailing power of hospitals.274 There were found to be a number of rivals in DI equipment and DPs that could undermine the ability of the merged entity to engage in foreclosure by employing counterstrategies such as price reductions, similar bundles (through teaming or counter-mergers) and technological leapfrogging as a result of innovation.275 As regards the incentive to engage in and effect of foreclosure, the parties put forward a number of arguments as to why marginalisation of rivals was unlikely with respect to both DI equipment and DPs. Before considering the specifics of this assessment in GE/Amersham, it should be noted that in many cases in which an analysis of the likelihood of foreclosure is required, a conceptual distinction can be made between foreclosure of existing and future products. Foreclosure of existing products occurs if, through reducing rivals’ sales, these rivals become marginalised and cease to exercise an effective competitive constraint with their existing or current product offering. For example, rivals may decide to withdraw from a market with their current product offering or find that their variable costs have increased, causing them to increase prices. Foreclosure of future products occurs if, through reduction in investments such as R&D or capacity, rivals become less able to exercise a competitive constraint with their future products than would otherwise have been the case. If there is no evidence that foreclosure will reduce the competitiveness of rivals with respect to their current products, the analysis inevitably becomes more speculative in nature, as it seeks to predict the effect on the competitiveness of rivals within the medium to long-term future. In GE/Amersham there was an important distinction to be made between existing and future products. With respect to existing products, the parties first noted that products would only be withdrawn from a member state if the costs that would be saved exceeded the expected forgone revenue. The production of both DI equipment and DPs was characterised by significant fixed costs, many of which (e.g. sunk R&D expenditure at a global level) would not be recovered or avoided even if production within a given member state were to cease entirely. Given the
274
Ibid., para. 38.
275
Ibid., para. 39.
conglomerate mergers
491
high contribution margins on sales of these products, the impact of any bundling strategy would have to reduce rivals’ margins significantly before they found it unattractive to remain present in a particular member state due to an inability to make incremental sales that contributed to overall profits. Second, unit variable costs were relatively constant and independent of the level of output produced, providing production did not become capacity constrained. This implied that even if rivals experienced a significant reduction in volumes, there would be no reason to believe that the competitive constraint they exerted on the merged entity for incremental business would be diminished. Third, with respect to DI equipment there was an ongoing requirement to service the existing installed base in each member state. To earn revenues from upgrades and servicing imaging equipment that had already been sold to customers would require a presence at the national level. Since the merged entity could not affect this source of revenue, it would be difficult to induce the exit of rivals from a particular country. Moreover, with respect to DPs, the merged entity would be unable to affect rivals’ sales to hospitals that already have in place DI equipment that they were not considering replacing in the near future since these customers would not consider the commercial bundle of the merged entity. Fourth, re-entering a particular member state would be both easy and timely as the costs of re-establishing a distribution network were low. This implied that even if rivals were to exit a particular member state, the merged entity would nonetheless remain unable to exploit a position of market power. With respect to future products, the parties noted that the incentives of rivals to invest in R&D would not be adversely affected, mainly because these incentives were not related to revenues generated in particular member states but to worldwide sales prospects. The Commission considered that anti-competitive foreclosure could arise only if rivals were forced to exit the DP and DI markets at a global level, and that any erosion of market share in certain member states would be unlikely to lead to such global exit. It also excluded the possibility that rivals would significantly reduce their global R&D expenditure. Finally the Commission accepted that barriers to entry or re-entry in a particular member state are not significant.276 As a result, 276
See Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004, paras. 40 and 41.
492
non-horizontal mergers
the Commission concluded that whilst the commercial bundling of DI equipment and DPs may occur post-merger, it would not lead to competitive concerns. 5.5.2.2 Technical tying As noted above, technical tying refers to a situation in which the post-merger entity’s products would work optimally or exclusively when used in conjunction with one another. At the time of transaction, all DI equipment worked with all available DPs. The Commission investigated the hypothetical concern, raised by a third party, that the merged entity would engage in technical tying that would ultimately foreclose rivals by denying them interconnectivity with the merged entity’s products.277 The Commission recognised that this concern could only be relevant in principle to new products since current products worked optimally with each other and no action could undo this compatibility. The Commission considered the technical links between the manufacturing of DI equipment and the development of DPs to investigate whether new Amersham DPs could be tailored to function exclusively or optimally with GE equipment.278 Its market investigation found that both customers and competitors found no interoperability issues with existing products. It found that to the extent that Amersham could develop breakthrough DP products that no other competitor could replicate and that could not be used with non-GE DI equipment, this would most likely take place in the field of nuclear imaging.279 However, after a detailed investigation it concluded that Amersham’s development plans in DPs, and in nuclear imaging DPs, did not raise concerns regarding the possibility that any product would work most effectively with future GE equipment to the exclusion of rival DI equipment.280 277
278
279
280
The Commission also considered a second related concern that the merged entity would acquire a ‘time-to-market advantage’ by internalising the knowledge on development plans in DI equipment and DPs carried out by each of the merging parties. However, it dismissed this concern on the basis of serious questions regarding the feasibility of such a practice and on the basis that any such advantage could only be short-lived. See ibid., paras. 56 and 57. Another case in which the Commission considered surrounding the practical ability to engage in technical tying is Saint Gobain/BPB. See Case COMP/M.3943 – Saint-Gobain/ BPB, Commission decision of 9 November 2005, para. 61. Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004, para. 49. Ibid., para. 54.
conglomerate mergers
493
The Commission also considered the incentive to engage in technical tying. It considered that the hypothesised strategy would deny the merged entity significant sales of Amersham’s products to the installed base of competing DI equipment. Given that rival DI equipment represented a substantial part of the nuclear imaging market due to GE’s relatively modest market shares in this area, the Commission concluded that there would be no incentive to pursue technical tying in the area in which it was considered most feasible.281 As a result of the above, the Commission dismissed the possibility of competitive concerns related to technical tying of DI equipment and DPs.
5.5.3 Google/DoubleClick In March 2008, the Commission cleared Google’s proposed acquisition of DoubleClick, a leading online advertising technology company.282 The main actors in the online advertising industry are publishers, who sell advertising space on their websites, and advertisers, who buy advertising space in order to reach internet users. Defining the relevant product market in which online ad space is sold is a complex task. First, online advertising appears in a variety of different formats. Google predominantly sold simple text ads, but the industry also includes static graphic ads and more advanced rich media ads (collectively referred to as display ads).283 Second, online advertising can either 281
282
283
See Case COMP/M.3304 – GE/Amersham, Commission decision of 21 January 2004, para. 59. Similarly, in IBM/Telelogic, the Commission considered whether the merged entity would have an incentive to engage in technical tying by withholding interoperability information, in order to leverage its market power in the markets for Modelling and Requirements Management tools into adjacent markets. It noted that the merged entity would incur substantial losses if it were to make its products incompatible with competing products. By comparison, it considered the benefits to the merger in terms of increased sales of tied products to be limited. See Case COMP/M.4747 – IBM/Telelogic, Commission decision of 5 March 2008, para. 268. See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008. In addition to exclusionary concerns the Commission also considered ‘diagonal effects’ concerns put forward by third parties. Diagonal mergers are discussed in Section 6 below. Much of the discussion of this case is based on S. Lewis and A. Lofaro, ‘Google/ DoubleClick: The search for a theory of harm’ (2008) 29(12) European Competition Law Review 717. Online ads are also differentiated by the way in which they are targeted. For example, an ad can be targeted according to a search query entered by a user (search ads), according to the content on the page on which it appears (contextual targeting), according to the user’s past viewing behaviour (behavioural targeting) or according to various other
494
non-horizontal mergers
be sold directly by a publisher to an advertiser or indirectly through ad networks or ad exchanges, which act as two-sided intermediation platforms matching advertisers with publishers providing suitable ad space. For example, Google sells ad space on the websites of third party publishers that make use of its ad network, AdSense. The parties were of the opinion, as were most of the complainants to the merger, that ads appearing in different formats, targeted by different methods and sold through different channels are all substitutable and therefore that the relevant market should at least comprise the sale of all online ad space. The Commission agreed that different types of online advertising are substitutable to a certain extent although it did not reach a firm conclusion as to whether the substitutability was sufficient to justify a single relevant market for all types of online advertising.284 Once a publisher has agreed to sell advertising space on its website to an advertiser (either directly or indirectly), ad serving technology is used to deliver an ad from the advertiser to the ad space, as well as playing various other important supporting roles on behalf of both advertisers and publishers.285 Since ad serving is an input into the provision of ad space, suppliers of advertising space and suppliers of ad serving solutions sell complementary products and therefore do not compete in the same relevant markets. Importantly, although all online ads, of any format, require some ad serving technology to place them on a website, the basic technology used to serve text ads is not substitutable for the more advanced technology, such as that provided by DoubleClick, used to serve display ads. This implies that there was no straightforward horizontal overlap between Google – a seller of ad space along with integrated (basic) ad serving for text ads, and DoubleClick – a seller of stand-alone display ad serving solutions. Complainants raised a number of non-horizontal theories of harm, including an allegation that the merged entity would leverage DoubleClick’s market position in ad display serving to foreclose ad intermediaries that compete with Google’s ad network (discussed in Section 5.5.3.1
284
285
indicators such as the user’s geographic location or the time of day the user is viewing the website. The Commission did conclude, however, that a separate market could be defined for the provision of ad intermediation. DoubleClick provides services to advertisers via its advertiser-side ad serving solution (DFA) and to publishers through its publisher-side ad serving solution (DFP).
conglomerate mergers
495
below). Another allegation was that the merged entity would leverage Google’s position in advertising and ad intermediation markets to foreclose display ad serving providers that compete with DoubleClick (discussed in Section 5.5.3.2 below). A further allegation was that a ‘diagonal relationship’ between Google and DoubleClick would make unilateral price increases profitable for the combined entity (discussed in Section 6 below). 5.5.3.1 Allegations regarding foreclosure of ad intermediation providers Complainants alleged that the merged entity would leverage DoubleClick’s market position in ad serving to foreclose ad intermediaries that compete with Google’s ad network. Whilst a wide range of concerns were put forward, they all involved the merged entity taking some action, such as increasing the price of display ad serving when used with competing networks, that would make DoubleClick’s advertiser and publisher customers favour Google’s ad networks over others.286 It was alleged that this would lead to the marginalisation of competing ad networks and that the competitive harm would be exacerbated by the existence of network effects arising from the two-sided nature of ad intermediation. The analysis of DoubleClick’s market power was an important part of the Commission’s assessment of these theories. The Commission confirmed the position set out in the Non-horizontal Merger Guidelines that substantial market power (but not necessarily dominance) is required for foreclosure concerns to arise. Specifically, the Commission stated: ‘As recognised by the Non-Horizontal Merger Guidelines, in order to be able to foreclose competitors, the new entity must have a significant degree of market power (which does not necessarily amount to dominance) in one of the markets concerned.’287 Importantly, the Commission’s decision makes clear that any assessment of the ability of a firm to foreclose its rivals will need to consider the extent to which the merged entity is constrained by rivals, and will need to go beyond a simple assessment of market shares in this regard.
286
287
In addition, the Commission considered a hypothetical strategy whereby DoubleClick would technically ‘tweak’ the ad arbitration mechanism to favour Google’s ad network, AdSense. See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008, para. 289. Ibid., para. 333.
496
non-horizontal mergers
Despite finding that DoubleClick had market shares of 30–40% in advertiser side ad serving and 40–50% in publisher side ad serving, the Commission recognised that DoubleClick was subject to substantial competitive constraints from rival ad serving providers that would undermine its ability to foreclose rivals. The Commission found that ad serving markets were competitive, on account of the fact that there was significant evidence of customer switching, evidence that DoubleClick’s customers were able to negotiate significant price (cost-per-thousand impressions or CPM) reductions at the point of renewing their contracts and evidence that prices for customers of all sizes were declining significantly over time despite increasing demand.288 For example, the Commission stated: The Commission considers that the evidence provided by the parties is convincing in showing that DoubleClick has had to reduce prices in response to competitive constraints by its rivals. As explained below, the parties’ data shows that CPMs have decreased within volume-tiers (and hence the fall in average CPM is not primarily driven by increases in volumes) and data on price reductions offered during renegotiations with specific customers at the time of contract renewals suggests that DoubleClick has responded to competitive pressures.289
The analysis submitted by the parties led the Commission to state in the context of assessing the merged entity’s ability to engage in foreclosure strategies based on DoubleClick’s position in ad serving that: ‘… the market investigation has revealed that DoubleClick faces a number of competitive constraints and is not able to exercise any significant market power.’290 The Commission’s conclusion on the absence of an ability to foreclose rival ad intermediation providers was further based on its view that DoubleClick’s customers did not face substantial switching costs.291 A key argument put forward by complainants was that DoubleClick’s ad serving price constituted a sufficiently high proportion of the cost of using an ad network that variation in this price could affect the relative attractiveness (to advertisers and publishers) of competing ad networks. However, as recognised by the Commission, the proportion of intermediation costs that ad serving represents is not a particularly relevant consideration for either publishers or advertisers when choosing between ad networks.
288 291
See ibid., para. 168. See ibid., para. 297.
289
Ibid., para. 169.
290
See ibid., para. 296.
conglomerate mergers
497
The relevant consideration for advertisers is the total cost of purchasing the ad space, and for publishers it is the total profit from selling such space. Even a relatively large increase in the price of ad serving when used with a competing ad network could not have a significant impact on these values and would therefore be very unlikely to influence advertisers’ or publishers’ choice of ad network.292 Moreover, competition within the markets for display ad serving would make any substantial price increase impossible. An important additional ingredient to the alleged foreclosure concern was that ad intermediation was prone to ‘tipping’ due to the existence of network effects that are often present in two-sided markets.293 In choosing to use a particular ad network, an advertiser may take into account the number of publishers using the same network, in addition to the price the advertiser must pay to use the network. Similarly, a publisher’s decision to join an ad network may be influenced by the number of advertisers using the network, as well as by the price it must pay to use the network (or its share in the advertising revenue generated). Complainants alleged that, using the various leveraging strategies available, the merged entity would attract so many advertisers and publishers to its ad network that other ad networks with fewer members would be marginalised. The merged entity would subsequently be able to raise prices without the constraints imposed by rival networks making such a price rise unprofitable, because other networks would be of lower ‘quality’ from the perspective of advertisers (publishers) due to a lack of publishers (advertisers) on the other side of the market. The concern that network effects may lead to tipping/monopolisation has been raised in other cases, most notably in relation to Microsoft’s Windows operating system, but such allegations must be carefully tested in each market context where they are made. The risk of tipping is
292
293
For example, suppose that a publisher opts for a non-AdSense network and sells a $2 per-thousand-impression ad, pays 40 cents to the ad network, and pays 5 cents to the publisher-side ad serving provider. The publisher’s net profits are $1.55 per thousand impressions. Now assume the price of publisher-side ad serving were to increase by 10% from 5 cents to 5.5 cents. If the publisher were to continue using the rival network to sell the ad space, its net profits would fall by only around 0.3%, from $1.55 to $1.545 per thousand impressions. Network effects may arise when consumer utility in a certain market depends on consumption of the same good or service by other agents. Two-sided markets are characterised by a particular type of network externality whereby the externality depends on consumption of ‘compatible’ agents on the opposite side of the market.
498
non-horizontal mergers
greater when customers typically use only one platform, and when network effects are not exhausted at a low level of usage. The market evidence supplied by the parties convinced the Commission that neither of these elements was present in the online advertising industry. First, advertisers and publishers frequently use more than one platform (i.e. they ‘multi-home’). Given that multi-homing is attractive and costless, it is unlikely that any network effects that may exist could give rise to anti-competitive effects in the form of foreclosure of rival ad networks. This is because the attraction of an advertiser or publisher to an ad network does not make that advertiser or publisher unavailable to other networks. Second, the evidence suggested that beyond a certain number of publishers an additional publisher joining the network provides no further benefit to advertisers already on the network. This is suggested by the observation that surveyed advertisers did not see reach (the number of unique monthly visitors to ad networks’ sites as a percentage of the online population) as a key differentiating factor between ad networks despite significant variation in reach across the various alternative ad networks in the market. Hence, a rival ad network may continue to exert a competitive constraint with a relatively small number of partners on the publisher side as long as it is attractive in terms of the various other dimensions along which ad networks are judged (e.g. the targeting method used). The Commission concluded that given the presence of credible alternatives to which DoubleClick’s customers can switch and the evidence that network effects are not strong enough to induce tipping, any strategy to attract publishers/advertisers to AdSense was unlikely to be able to foreclose rivals in intermediation markets.294 Regarding the incentive to foreclose, the Commission considered that pure bundling of ad serving with ad intermediation would be likely to be unprofitable.295 It noted that while the margins earned on additional sales of the bundle would exceed those lost on sales of the ad serving tools, the strategy was likely to be unprofitable in view of the switching it would entail to rival ad serving products.296 The Commission also considered the incentive of the merged entity to engage in mixed
294
295
See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008, para. 310. 296 See ibid., para. 315. See ibid., para. 315.
conglomerate mergers
499
bundling (rendering ad serving products more expensive or lower quality when used with other ad networks). The Commission noted that publishers that were customers of DoubleClick’s ad serving products spend relatively limited amounts on Google’s non-search platform and that incentives to offer a mixed bundle might be limited as margins earned on additional sales through AdSense would merely (if at all) compensate the opportunity cost of reducing the price of DoubleClick tools.297 Finally, regarding the effect of foreclosure, the Commission noted that anti-competitive foreclosure was unlikely to arise because the new entity would continue to compete with a number of platforms offering the same product combination. Specifically it noted that: As a result of a recent wave of acquisitions, the market has evolved to a situation where ‘bundled’ platforms (intermediation + ad serving tools) now coexist with independent suppliers of inputs for online advertising (ad networks and ad exchange offering intermediation only and suppliers of standalone ad serving tools). In response to bundling strategies by the new entity, the competing platforms could respond by offering similar bundles.298
The ability of rivals to match a bundling strategy by the merged entity has been seen as an important factor that undermines the potential for anti-competitive foreclosure to arise in a number of other cases, including Johnson&Johnson/Guidant,299 GE/Smiths Aerospace,300 and Procter&Gamble/Gillette.301 297
298 299
300
301
See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008, para. 314. Ibid., para. 327. In Johnson&Johnson/Guidant, the Commission was satisfied that bundling would not give rise to anti-competitive foreclosure of rivals, because a bundling strategy could be matched by a number of competitors: Case COMP/M.3687 – Johnson&Johnson/Guidant, Commission decision of 25 August 2005, para. 341. In GE/Smiths Aerospace, the Commission noted that a competitor of Smiths Aerospace, UTC, was able to offer engines and power generation systems so that it would be able to replicate any bundling of these products by the merged entity and that Smiths Aerospace’s other competitors would be able to respond to the tying behaviour of the new entity by teaming up with another engine manufacturer: Case COMP/M.4561 – GE/ Smiths/Aerospace, Commission decision of 23 April 2007, para. 124. In Procter&Gamble/Gillette, although the merger would create greater opportunities for bundling, anti-competitive foreclosure was considered unlikely to arise given the presence of strong rivals with similar portfolios and powerful buyers. Importantly, the Commission accepted that Procter&Gamble/Gillette’s enlarged product range may lead to efficiencies from one-stop-shopping and economies of scope (e.g. lower logistic costs). See Case COMP/M.3732 – Procter&Gamble/Gillette, Commission decision of 15 July 2005, paras. 121–2 and 131.
500
non-horizontal mergers
5.5.3.2 Allegations regarding foreclosure of ad serving providers Complainants alleged that the merged entity might use Google’s market position in search advertising and (search) ad intermediation to foreclose DoubleClick’s rivals by bundling its sales of search ads or its intermediation services for the sale of search and/or non-search ads with DoubleClick’s ad serving technology. The Commission considered both pure bundling and mixed bundling strategies. Under pure bundling, advertisers wanting to place search ads via Google’s search advertising solution (AdWords) would be (contractually) required to make a certain minimum use of DFA, and/or publishers wanting to use Google’s (search) ad intermediation solution (AdSense) would be obliged to use DFP, either on a contractual basis or by means of a technological tie. Under mixed bundling, Google would offer conditional discounts to induce advertisers and publishers who utilise AdWords or AdSense to (voluntarily) utilise DFA or DFP. The Commission concluded that while the ability to engage in these forms of foreclosure could not be entirely excluded, the merged entity would be unlikely to have the incentive to engage in foreclosure and that, in any event, such a strategy would not have a significant detrimental effect on competition because a number of financially strong, vertically integrated competitors would not be foreclosed.302 Regarding ability to foreclose, the Commission noted in particular that there would be practical difficulties in requiring advertisers that want to place search ads via Google’s AdWords product to use DoubleClick’s ad serving product DFA. This is because prices for search advertising are determined by an auction on a continuous basis (every time a user enters a search query) whereas the terms according to which DoubleClick provides display ad serving are set by contracts that typically have a duration of one to two years, making it difficult to set the terms for these products simultaneously.303 These practical difficulties effectively enabled the Commission to rule out bundling on the advertiser side. Nonetheless, the Commission found that on the publisher side, the practical difficulties in bundling Google’s (search) ad intermediation with DFP appeared to be more limited because for the provision of both
302
303
See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008, para. 332. See ibid., para. 340.
conglomerate mergers
501
display ad serving and (search) ad intermediation for (larger) publishers, contractual arrangements of a similar nature and duration applied.304 However, the Commission noted that the common pool of customers of Google and DoubleClick is currently fairly limited, reducing the ability of the merged entity to foreclose.305 Regarding the incentive to foreclose, the Commission found that bundling Google’s sales of search ads and (search) intermediation services with DFA and/or DFP would not be profitable. It noted that as margins on DFA and DFP are low compared to margins on Google’s direct sales of search ads and intermediated sales of (search) ads, even small volume losses in search advertising and (search) intermediation (due to customers rejecting the bundle) would outweigh the gain in profits from customers taking up DFA or DFP.306 The Commission also recognised that, as discussed in Section 5.4.2 above, an important benefit to bundling disappears if prices are individualised, rather than uniform across customers. In particular, the Commission noted: [I]n the online advertising environment transactions often involve customised solutions or services that are uniquely priced. Bundling is usually an attractive and profitable strategy in order to discern customer’s willingness to pay in a context where prices are posted and uniform across customers. In the online advertising industry, bundling would not enable the new entity to increase profits because prices are highly individualized. On the advertiser side, both Google through its auction mechanism for keywords and DoubleClick through its direct negotiation with customers have the ability to vary the price of their products according to customers’ willingness to pay. Similarly, on the publisher side, both Google through its negotiations with its direct partners (which account for around [>80%] of its AFS revenue) and DoubleClick through its direct negotiation with publishers, have the ability to vary the price of their products according to publishers’ preferences. In such context, one of the attractions of bundling usually disappears.307
Finally, regarding the effect of foreclosure, the Commission reiterated that the merged entity would be very unlikely to marginalise competitors such as Microsoft, Yahoo! and AOL, noting that each of these competitors was vertically integrated and had access to considerable financial resources.308
304 307
See ibid., para. 341. Ibid., para. 353.
305 308
See ibid., para. 343. See ibid., para. 357.
306
See ibid., para. 348.
502
non-horizontal mergers
6.
Diagonal mergers
The term ‘diagonal merger’ refers to a particular type of non-horizontal merger in which the merging parties are an upstream supplier and a downstream competitor of the firms for which the upstream supplier’s good is an input.309 The merger is not horizontal because the merging parties are not active in the same relevant market. Moreover, the merger is not vertical because the good produced by the upstream supplier is not an input as far as the downstream firm is concerned, despite it being an input for some of the downstream firm’s competitors. Whilst there is no customer that would switch from the good produced by one merging party to a good produced by the other merging party in response to a price increase, the two products may be substitutable in an indirect sense. This indirect substitutability arises because an increase in the price the upstream supplier charges to rivals of the other merging party will increase the costs of those rivals, potentially causing them to increase their prices, diverting downstream demand to the other merging party. A diagonal merger can therefore in principle give rise to a unilateral incentive to increase price. Equally, a diagonal merger may give the downstream firm an incentive to increase its price if such a price increase would increase its rivals’ sales and hence increase demand for the input supplied by the upstream merging party. In assessing the extent to which a diagonal merger could give rise to an incentive to increase price it is necessary to consider: • whether the input produced by the upstream firm is an important input for rivals of the downstream merging party; • whether the upstream firm faces effective competition from other firms producing the input; and • whether the products produced by downstream rivals are close substitutes for the products produced by the downstream merging party. To understand further the unilateral effects concerns associated with diagonal mergers and the importance of each of the above considerations it is instructive to consider a stylised example in which a diagonal relationship could be said to arise. 309
This is equivalent in economic terms to a merger between a Firm A and a competitor of Firm B, whereby Firm A supplies a product that is a complement to the product supplied by Firm B.
diagonal mergers
Copper
Zinc
503
Iron
Brass
Figure 5.13
Coal etc.
Steel
Illustration of a diagonal merger
6.1
Diagonal mergers: an example
This section presents a hypothetical example of a diagonal relationship between a supplier of zinc and a supplier of steel.310 Zinc and copper are complementary inputs to the production of brass. Brass is a substitute for steel in certain applications. There are no customers that would consider switching from zinc to steel in the event of an increase in the price of zinc. However, these two products may nonetheless be indirectly substitutable: an increase in the price of zinc that results in an increase in the price of brass would cause an increase in the demand for steel; and an increase in the price of steel would cause an increase in the demand for brass, and therefore for zinc. A merger between a steel provider and a zinc provider would internalise this relationship, potentially giving rise to an incentive to increase the price of zinc and/or steel. Figure 5.13 illustrates the diagonal relationship between zinc and steel. The following three conditions would need to be satisfied in order for the unilateral effects concern discussed above to materialise: • Zinc must represent an important cost in the production of brass. To see why this is the case, suppose that zinc represented only 10% of the cost of producing any given volume of brass. In that case, a 10% increase in price by the zinc supplier would give rise to only a 1% increase in the production cost of brass. This would be unlikely to give rise to a
310
Higgins puts forward the example of a merger between a zinc supplier and steel supplier in the article that introduced the term ‘diagonal merger’. See R. S. Higgins, ‘Diagonal Merger’ (1997) 12 Review of Industrial Organization 609–23.
504
non-horizontal mergers
significant increase in the price of brass and thus would be unlikely to give rise to a significant increase in the demand for steel. In effect, the low proportion of the cost of zinc in the overall cost of producing brass serves to lower the cross-price elasticity of demand for steel with respect to the price of zinc. • The zinc provider should not face effective competition from other zinc providers. Suppose instead that the zinc supplier faced many competitors in the supply of zinc to brass suppliers. In that case, any increase in price by the zinc supplier will give rise to substitution to rival zinc suppliers by brass-producing customers, rather than any increase in the production cost of brass. • Brass and steel should be close substitutes. If brass and steel were not close substitutes, then even a significant increase in the price of brass would not cause much of an increase in the demand for steel produced by the other merging party. In effect, the low degree of substitutability between brass and steel serves to lower the cross-price elasticity of demand for steel with respect to the price of zinc. Each of these conditions must be considered simultaneously. In particular, the higher the degree of substitutability between brass and steel, the lower in principle need be the importance of zinc as a proportion of the cost of producing brass for a significant substitutability relationship to exist.
6.2
Case study: Google/Doubleclick – assessment of diagonal effects
The Commission had to consider the potential for a diagonal relationship to give rise to unilateral effects concerns during its investigation into the merger between Google and DoubleClick.311 As discussed in Section 5.5.3 above, there was no straightforward horizontal overlap between Google – a seller of ad space along with integrated (basic) ad serving for text ads, and DoubleClick – a seller of stand-alone display ad serving solutions. The absence of horizontal overlaps between the parties’ operations did not, however, rule out the possibility that the post-merger entity could have an immediate incentive to raise prices unilaterally. Both the parties and complainants to the merger agreed that advertisers and publishers see text advertising and display advertising as 311
See Case COMP/M.4731 – Google/DoubleClick, Commission decision of 11 March 2008.
diagonal mergers
505
substitutes. Complainants alleged that this created a diagonal relationship between Google and DoubleClick of the form described in the example above that would make unilateral price increases profitable for a combined entity. Complainants alleged that an increase in the price of DoubleClick’s advertiser-side ad serving solution (zinc in the example above) would increase the total cost to an advertiser of purchasing display advertising space (brass) in the unintegrated channel, to which DoubleClick’s product is an input. Since the unintegrated channel is viewed as substitutable for the integrated channel in which Google sells text advertising (steel), there would be some diversion of demand to this channel, and some diversion of demand to Google. This diversion would be internalised by a combined entity, leading to an incentive to increase the price of the display ad serving solution.312 Whilst this concern makes sense in theory, analysis conducted by the parties revealed that none of the three conditions described in the context of our stylised example was present. First, in choosing between different forms of advertising (e.g. text and display), advertisers consider the total cost of one form versus the total cost of another. Advertisers might well respond to a small but significant (5–10%) increase in the total cost of display advertising, for example, by re-allocating expenditure to text advertising. However, since display ad serving constitutes a small proportion of the total cost to the advertiser of display advertising, small but significant changes in the price of display ad serving can cause only tiny changes in the total cost of display advertising relative to the total cost of text advertising. To take a concrete example, suppose an advertiser pays 5 cents per thousand impressions in display ad serving fees to DoubleClick and pays $2 per thousand impressions to the publisher for the purchase of the ad space on the website. The total cost of advertising is therefore $2.05. If the price of the advertiser tool increases by 10% (from 5 cents to 5.5 cents), this would raise the total cost of advertising by only around 0.2% (from $2.05 to $2.055). As pointed out by the Commission, such price changes are therefore very unlikely to precipitate much (if any) switching from display advertising to text advertising. 312
Similarly, an increase in the price of DoubleClick’s publisher-side ad serving solution (DFP) would reduce the profits to a publisher from selling ad space in the unintegrated channel, leading to an increase in the amount of ad space sold in the integrated channel – an effect that would also be internalised by the combined entity.
506
non-horizontal mergers
Second, DoubleClick faced strong competition within the markets for display ad serving. As noted in Section 5.5.3.1, for some time, DoubleClick had been forced to offer advertisers and publishers large price reductions at the point of renewing their contracts and, in addition, had lost a significant number of customers to rival display ad serving providers despite offering equally large price reductions. Competition within display ad serving implied that DoubleClick could not bring about a small increase in the price of ad serving, let alone one sufficient in size to cause any material switching to text advertising. Third, Google supplies text ads whereas DoubleClick is an input provider to the supply of display ads. While these forms of advertising are substitutable, they are clearly differentiated to some extent. Therefore, it is unlikely that Google’s integrated solution and an unintegrated solution that included DoubleClick’s display ad serving technology as a component could be considered as particularly close substitutes. The considerations set out above convinced the Commission that Google and DoubleClick did not exert a significant competitive constraint on one another and therefore that no unilateral price increase would be profitable post-merger.
Appendix A Regression analysis and econometrics
This Appendix provides a brief explanation of linear regression analysis and its use in econometrics. The goal of Section 1 is to allow the reader to gain a basic understanding of the output of regression analyses. In Section 2, we provide a brief introduction to a fundamental topic in econometrics: the identification of economically meaningful relationships. It is not our intention to provide a detailed guide to carrying out regression analysis in practice, a topic that is beyond the scope of this book. Moreover, we do not discuss the myriad of issues and complications that will often arise in regression analysis. A more detailed treatment of this topic can be found in any standard econometrics textbook.1
1. Regression analysis and statistical inference Regression analysis is a tool for identifying a mathematical description reflecting an observed relationship between two or more variables. Regression analysis can allow a researcher to understand how the value of one variable tends to change when one or more other variables change. To give a simplified example, consider the pricing of television sets. Let us imagine that a TV producer uses a ‘rule of thumb’ to set prices for its 3D ready, LED flat screen TVs. This pricing rule is based on only one aspect of the TV: its screen size. The rule can be expressed as a mathematical formula that will provide the price of a TV of any given screen size. Specifically, the TV price is given by the following equation, where P is price, and SCR is screen size in inches, minus 26 (26 inches being the minimum screen size available): P ¼ 200 + 68SCR This expression tells us that the price, in Euros, will be equal to the screen size over and above the minimum 26 inches multiplied by €68, plus €200. 1
See, e.g., J. M. Wooldridge, Introductory Econometrics: A Modern Approach, 4th edn (Mason, OH: Cengage Learning, 2009).
507
508
regression analysis and econometrics
Price ( )
2,512
1,152
200 0 26
40
60 Screen Size (”)
Appendix A – Figure 1
Illustration of rule for setting price given screen size
Thus, for any given screen size, the TV producer uses this equation to set the price. A 26-inch TV will have a price of €200, equal to 200 + 68 (26 – 26). A 40-inch TV will have a price of €1,152, equal to 200 + 68 (40–26). A 60-inch TV will have a price of €2,512, equal to 200 + 68 (60–26). The relationship between screen size and price embodied in this equation can also be expressed graphically, as shown in Figure 1. For any given screen size, the line plotted in the figure shows the price in Euros. In this example, the producer is using a known value of one variable (screen size) to set the value of another (price). Regression analysis reverses this process: rather than starting with the formula describing the relationship between variables to predict the value of one variable from the other, regression analysis starts with observed values for the variables and attempts to predict the formula that governs their relationship. Suppose now that we, as researchers, hypothesise that the TV producer uses a fixed rule to set TV prices but that the specification of that pricing formula is not known to us. Note that we are not interested in understanding the economic mechanisms that cause the producer to set higher
regression analysis and statistical inference
509
3,000
2,500
Price ( )
2,000
1,500
1,000
500
0 26
31
36
41
46
51
56
61
Screen size (”)
Appendix A – Figure 2 Illustrative regression – predicting prices from screen size
prices for larger TVs (such as the possibility that larger TVs have higher marginal costs of production, or that consumers are willing to pay more for larger TVs). We are only interested in uncovering the rule that the producer uses to set price. Suppose that we have data showing the price and screen size of eighteen television sets sold by the producer, all with the same specification other than size (i.e. they are all 3D ready, flat screen with LED technology). The price and screen size of these eighteen observations are plotted in Figure 2. Looking at these eighteen observations, the relationship between price and screen size for this product is relatively clear. While it is not possible to draw a straight line that goes through all the observations, the line shown on the chart can be seen to be a good reflection of the relationship. Regression analysis is a mathematical tool for identifying the equation that describes the so-called ‘line of best fit’ for a particular set of observations. In most economic applications, we are interested in the effect of a change in an explanatory variable (here, screen size) on a dependent variable (here, price). That is what regression coefficients provide.
510
regression analysis and econometrics
In this case, the regression coefficient represents an estimate of how much we can expect the producer to increase the price of a 3D ready, LED flat screen TV if the TV has a one inch bigger screen. Running a regression on the above data finds a coefficient for screen size of 68, which corresponds to the TV pricing formula described above. The dotted line above is an attempt to describe the effect of screen size on the price of a TV, but none of the observations actually lies on that line. So we could depict the real relationship (the one we actually observe) as: P ¼ 200 + 68SCR + SOMETHING ELSE This ‘something else’ is referred to as the unobservable or the error term; it is something that affects the economic process that we as researchers cannot see and therefore cannot incorporate into our model. In Figure 2, this error represents the distance between the dotted line and the data point, and is called the residual error, or residual for short. We often have an idea, whether through application of economic theory or industry knowledge, of what factors might lie within the error term. For instance, one thing that we did not account for in our analysis was the ‘newness’ of the television set; each year when a new design of television sets comes out, older versions may still be sold, but at a lower price. It is likely that newer models look more stylish, are thinner and lighter or have a smaller stand. These things will affect the price of the set but are not included in our model. As discussed further below, we can include additional variables in a regression that explain some of the variance in price and thus allow us to identify the effect of screen size more accurately. There are various forms of regression procedure for identifying the equation of a line that best fits the observed data. The most common technique is ordinary least squares (‘OLS’, the technique used here), which operates by minimising the sum of the squared residuals. The above example is a very specific regression, in which we ask ‘for 3D ready, LED flat screen TVs, what is the relationship between screen size and the price the producer sets?’ If we were interested instead in the effect of screen size on the price of flat screen TVs more generally, then we would need to include more observations to account for the fact that TVs are also differentiated according to their screen technology (e.g. LED or Plasma) and other variables. In practice, regression analyses can readily be expanded to account for multiple explanatory variables. Thus, in the context of the example
regression analysis and statistical inference
511
above, the dependent variable of TV price might be explained not only by screen size, but also by screen type (LED or Plasma technology), whether it is ‘3D ready’, whether it has Wi-Fi built in, its release date and other factors. Although such multi-dimensional relationships cannot be drawn graphically as in the two variable cases discussed above (such a chart would need to contain one dimension for each explanatory variable plus one for the dependent variable), the principle remains the same. Thus, a multivariate regression analysis will produce an equation that minimises the sum of square residuals from the observed combinations of the various variables included in the regression. Suppose we applied the simple regression above (price on screen size) to a larger, more varied, sample that included TVs of various different specifications. This sample has forty-two observations and includes TVs that differ in terms of their screen technology, series model age, whether the TV is 3D ready and whether the TV is Wi-Fi enabled, as well as their screen size. Table 1 shows the illustrative output from multivariate regression analysis that includes a number of explanatory variables in addition to screen size. While the format of the output will vary according to the software used, the key elements for the interpretation of the regression results will be common to all software packages.2 Table 1 presents the results for a regression that relates observations on the price of TVs to the following variables: the screen size in inches; a so-called dummy variable, ‘LED’, that takes the value 1 if the TV has an LED screen and 0 if it has a Plasma screen; a dummy variable, ‘3D’, that takes the value 1 if the TV is 3D ready and 0 otherwise; a dummy variable, ‘Wi-Fi’, that takes the value 1 if the TV is Wi-Fi enabled and 0 otherwise; and a dummy for each of the model series available (if all are zero then the TV is the most recent model). These variables have been chosen on the hypothesis that the price of a television will increase with higher specifications and newer models. The core element of any regression results is the set of coefficients. These are the estimated values of the gradient of the regression equation relating each explanatory variable (screen size, screen technology, series 2
The process of using regression analysis to calculate coefficient estimates can be arithmetically burdensome, but can readily be performed by statistical analysis software packages, such as Stata, E-Views or SAS. Indeed, Microsoft Excel is capable of performing multivariate OLS regression analysis, albeit in a slightly cumbersome fashion, using the ‘linest’ command.
512
regression analysis and econometrics
Table 1 Illustrative OLS regression, dependent variable shipment price (€) Standard 95% confidence Coefficient error t-statistic p-value interval Screen size LED 3D Wi-Fi Last model 2 models ago 3 models ago 4 models ago Constant n R2
56.13 427.02 50.03 145.51 244.71 592.35 953.20 804.39 309.2334 42 0.907
5.56 127.54 136.12 113.17 165.58 115.44 157.39 179.91 271.7705
10.10 3.35 0.37 1.29 1.48 5.13 6.06 4.47 1.14
0.000 0.002 0.716 0.207 0.149 0.000 0.000 0.000 0.263
44.82 167.54 326.97 84.74 581.59 827.21 1273.42 1170.41 243.69
67.43 686.51 226.91 375.77 92.16 357.48 632.98 438.37 862.15
age, etc.) to the dependent variable (price). The coefficient on each variable tells us how the average value of the dependent variable changes with a unit change in that explanatory variable. The coefficient on screen size is 56.13, meaning that each increase in screen size of one inch is on average associated with a €56.13 increase in the price of a TV, all else being equal. The first point to review with a set of regression results is the sign, or direction, of the coefficients. In Table 1, the regression results show that the estimated coefficient on the LED dummy is positive. This indicates that, for a TV of a given screen size and model age, if the screen uses LED technology as opposed to Plasma we can expect the TV to be more expensive. The negative coefficient on the ‘3 models ago’ dummy indicates that we can expect to pay a lower price for a TV of a given size and technology three releases old than an equivalent TV of the most recent model. The sign or direction of coefficients must be interpreted in light of the question of statistical significance. Recall that we hypothesised that there is a ‘true’ underlying pricing rule that the producer employs to set prices for TVs. We are trying to identify that rule using a sample of forty-two observations of TV prices and characteristics. We observe that TVs that are Wi-Fi enabled have higher prices, on average, than those that are not (the sign of the coefficient on Wi-Fi is positive). However, this may just be an artefact of the forty-two TVs we happen to have sampled and may
regression analysis and statistical inference
513
not be a feature of the ‘true’ underlying pricing rule. The concept of statistical significance relates to the confidence with which we can say that an observed effect has not arisen purely by chance. A particular coefficient on an explanatory variable (e.g. screen technology) is said to be ‘statistically significant’ if we can say that there is a low probability that the coefficient on that variable is different from zero by chance and we are hence able to reject the hypothesis that the variable has ‘no effect’ with a certain degree of confidence.3 The statistical significance of reported coefficient estimates may be presented in one or more of four ways: standard errors, confidence intervals, t-statistics, and/or p-values. The example above presents all four of these. The extent to which we can be confident that there is a ‘true’ underlying effect will depend on the size of the effect we observe (i.e. the size of the coefficient) and on the degree of variation around that average effect we observe in the data. The smaller the variance around the average effect, the more likely that coefficient is to be an accurate reflection of the ‘true’ underlying effect of that variable. This variance is quantified via the standard error associated with each coefficient. Whatever our chosen model, the coefficient we arrive at is always an estimate of the true parameter of interest. There will therefore be a range of alternative values surrounding our coefficient that could feasibly be the true value. If our estimator is unbiased then, given our data set, the probability of a given value being the true value declines the farther away it is from the estimate. If the estimate of a coefficient is 427.02, as for the LED screen dummy in the example above, the true value is more likely to be 427.02 than 450, which in turn is more likely to be the correct value than 375. Once we know the standard error of an estimated coefficient, probability theory allows us to calculate the likelihood that the true value of the coefficient lies a particular distance from the estimate, given the data that we have in our sample. It is a standard result that there is a 95% probability that the true value of the coefficient lies within approximately two (1.96) standard errors of the estimated coefficient. Thus, in Table 1, there is a 95% probability that the coefficient on the screen being LED as opposed to plasma lies within the range from 427.02 plus and minus 1.96 times the standard error (i.e. 167.54 to 3
Often, in short hand, the explanatory variable itself is referred to as ‘statistically significant’.
514
regression analysis and econometrics
686.51). This range is called the confidence interval, and is shown in the final two columns of the table. In regression analysis, we are often concerned with the direction of relationships. The principal use of standard errors and confidence intervals is to determine the likelihood that an estimated coefficient is statistically different from zero. A coefficient is said to be statistically significant at the 5% level (or significantly different from zero at the 5% level) if there is a 95% probability that the true value lies on the same side of zero as the estimated coefficient. This is equivalent to examining whether the 95% confidence interval includes only numbers of the same sign as the estimated coefficient or crosses the value of zero to include values of the opposite sign. The standard approach in econometrics is to reject the hypothesis that a coefficient is zero (i.e. that the explanatory variable has no effect on the dependent variable) if the coefficient is statistically significant at the 5% level.4 For those coefficients that do not meet the specified level of statistical significance we cannot reject the hypothesis that they are equal to zero at the 5% level. For example, from the output above, we can reject the hypothesis that the effect of TV screen size on its price is zero at the 5% significance level, since the 95% confidence interval does not include zero. However, we cannot reject the hypothesis that the effect of the TV being 3D ready on its price is zero at the 5% significance level, since the 95% confidence interval does include zero.5 The t-statistic and the p-value are alternative ways of summarising the significance of regression coefficients. The t-statistic represents the ratio of the coefficient to the standard error. As such, the absolute value of the t-statistic reflects the number of standard errors each coefficient lies away from zero. As per the discussion above, probability theory tells us the coefficients within 1.96 standard errors of zero are insignificant at the 5% level; an absolute t-statistic of more than 1.96 therefore indicates that a coefficient is significant at the 5% significance level.6 4
5
6
The 5% threshold is arbitrary. Sometimes a significance level of 1% or 10% may be chosen, depending on our degree of aversion to the prospect of erroneously rejecting the ‘no effect’ hypothesis. We use the term ‘do not reject’, rather than ‘accept’, in hypothesis testing because the results of the test necessarily point towards the variable’s significance, but do not rule out all other possibilities the way that ‘to accept’ indicates. The fact that we cannot rule out that the size of an effect is zero, does not mean we can ‘accept’ that the size of an effect is (exactly) zero. The equivalent values for the 1% and 10% significance levels are 2.58 and 1.65, respectively.
endogeneity and the identification of economic effects 515
The p-value is related to the t-statistic, and represents the probability of observing a t-statistic that is as extreme as or more extreme than the one observed assuming that the hypothesis of ‘no effect’ is true. A p-value of less than 0.05 indicates that the coefficient is significant at the 5% level, and vice versa. Thus, while the t-statistic allows a coefficient to be evaluated against a specified significance level, the p-value identifies the critical significance level at which the coefficient would become statistically insignificant. The remaining elements of the regression output are the constant, the number of observations, and the coefficient of variation. The constant represents the intercept for the regression equation; that is, the value of the price when the value of the screen size variable is 26 inches (since we defined the variable as the size of the screen minus 26 inches) and each of the other variables included in the regression is equal to zero. In many cases, the constant term is of limited interest. The analyst is often interested in the extent of the relationship between the explanatory and dependent variables; that is, the effect of a change in an explanatory variable on the dependent variable, rather than the level of the dependent variable in and of itself. The number of observations, frequently denoted by the letter ‘n’, simply reflects the number of observed combinations of values used to estimate the regression equation. As noted above, this will affect the reliability of the estimates via their statistical significance. The coefficient of determination (or R2) is a statistic reflecting the ‘goodness of fit’ for the regression. The R2 statistic lies between 0 and 1 and measures the proportion of the variation in the dependent variable that is explained by the explanatory variables. An R2 of 1 means that the explanatory variables explain 100% of the variation in the dependent variable. An R2 of 0 means that the explanatory variables explain none of the variation in the dependent variable. However, while it can be a useful indicator of the goodness of fit of a regression, the R2 may not be of particular interest if our main concern is whether a particular variable (e.g. whether the TV is LED or Plasma) affects price holding other variables constant.
2.
Regression analysis, endogeneity and the identification of economic effects
The preceding discussion has introduced the concept of regression analysis, which provides a tool for identifying the mathematical
516
regression analysis and econometrics
relationship between observed variables. Regression analysis is a frequently used technique in the field of econometrics, which aims to link real-world data to economic theory in order to empirically identify economically meaningful relationships. In many cases, however, the identification of a statistical relationship between variables will fail correctly to represent the underlying economic process governing those variables that is of interest to the researcher due to the issue of endogeneity. In very loose terms, endogeneity can be defined as an issue which stops us inferring that an apparent statistical relationship between two variables can be equated with any meaningful economic relationship between those variables. In this section, we provide an illustration of endogeneity in order to demonstrate its potential scope to lead to regression results that give a distorted view of the true economic relationship of interest. Our example concerns a study of the effect of education on a person’s salary.7 We might observe that people with more years of education on average earn higher salaries. In particular, a regression analysis might find that, on average, one more year of education is associated with x more Euros per year of salary. This might suggest the conclusion that one more year of education will on average enable an individual to achieve a higher salary of x Euros per year. This approach, however, does not take account of the fact that individuals choose how many years of education to obtain. Many of the factors that influence that choice, such as individuals’ innate ability and work ethic are unobservable to the econometrician. Critically, some of the unobservable variables that influence educational choice are also likely to affect how much an individual earns independently of the amount of education they obtain. Therefore, the finding that people with more years of education earn more might be explained (at least in part) by the fact that individuals with a higher ability tend to choose more years of education. In this case the statistical relationship between years of education and salary does not tell us about the incremental effect of education on salary, which is the relationship of economic interest. Figure 3 is an illustration of how the statistical relationship between education and salary may lead
7
This is a well-known area of economic research in which endogeneity problems arise. See, e.g., Z. Griliches, ‘Estimating the Returns to Schooling: Some econometric problems’ (1977) 45 Econometrica 1–22.
endogeneity and the identification of economic effects 517 Salary Salary as function of Education observed in data Salary as function of Education for High Ability group
Salary as function of Education for Low Ability group
Years of Education
Appendix A – Figure 3 Relationship between salary and education
us to overstate the economically meaningful relationship that describes the additional impact on salary of an additional year of education. Figure 3 shows a number of observations of years of education and salary for individuals with ‘low ability’ (denoted by bold crosses) and a number of observations of years of education and salary for individuals with ‘high ability’ (denoted by light crosses). It can be seen that ‘low ability’ individuals on average choose lower levels of education than ‘high ability’ individuals and for any given level of education earn a lower salary. The ‘true’ underlying impact of an additional year of education on salary for each group is shown by the slope of a line of best fit through each of these sets of points. However, it is not possible in practice to observe which points are associated with ‘low ability’ individuals and which points are associated with ‘high ability’ individuals. Drawing an ‘observed’ line of best fit through all the points as a whole leads to a line that is steeper than the ‘true’ relationship between education and salary within either group. The slope of the ‘observed’ line of best fit does not therefore reveal the true impact of an additional year of education on salary.
518
regression analysis and econometrics
The problem of endogeneity discussed above would disappear entirely if any of the following were true: – we could clone individuals and send them off to work while keeping their clones in education another year. By construction, the only difference between each individual and his or her clone would be the amount of education they receive. In this way we could directly infer the incremental economic effect of a year of schooling simply by considering the average difference in salary between each individual and his or her clone; – a ‘mad dictator’ decided how much education each individual should obtain on a completely randomised basis. Recall that the endogeneity problem arises because the finding that people with more years of education earn more might be explained (at least in part) by the fact that individuals with a higher ability tend both to earn more for any given level of education and to choose to obtain more education. If educational choice is removed from the individuals, then the endogeneity problem disappears. This is the basis for randomised trials in medical research, which are organised such that the only difference between treatment and control groups is whether or not they receive the treatment, allowing any systematic difference in medical outcomes between the groups to be attributed purely to the treatment; – we could observe ‘innate ability’ directly. If we could observe ‘ability’ directly, we could simply divide all individuals into a ‘high ability group’ and a ‘low ability group’ and find the relationship between education and salary for each group. Unfortunately for economic researchers none of these solutions is realistic. In particular, econometricians rarely have the ability to carry out experiments under controlled conditions. As such, the field of econometrics has devoted particular attention to developing techniques to circumvent endogeneity. A common approach is to employ instrumental variables. A working definition of an instrumental variable is a variable that affects one part of a system but not another, thereby enabling the researcher to identify a relationship of interest.8 While the use of
8
More precisely, the main requirements for using an instrumental variable are that the instrumental variable must be correlated with the endogenous explanatory variables, conditional on the other covariates; and the instrument cannot be correlated with the error term in the explanatory equation.
endogeneity and the identification of economic effects 519
instrumental variables is beyond the scope of this book, a more detailed treatment of this technique can be found in any standard econometrics textbook.9 9
For a further discussion of endogeneity and instrumental variables, see J. M. Wooldridge, Introductory Econometrics: A Modern Approach, 4th edn (Mason, OH: Cengage Learning, 2009).
Appendix B Models for demand estimation
1.
Introduction
In Chapter 2 we considered a simple form of demand estimation based on linear demand. This appendix provides some further background on different forms of demand estimation that are applied in the context of merger investigations.
1.1 Alternative models for continuous demand estimation Consider the linear illustrative demand curve for Jaffa orange juice presented in Chapter 2, Section 3.2.1.1 at Figure 2.6. From this demand curve we can see that a one unit increase in price will lead to a fall in demand of 10 units, a relationship that holds true at all prices. Consequently, in using a linear demand system we have imposed the restriction that if the price falls from £10 to £0, for example, we will have the same increase in demand as if the price had fallen from £150 to £140. Intuitively this does not seem like a desirable attribute for a demand curve to have. It is possible to specify a demand relationship that does not have this property. Consider a demand function that is described by the following equation: β
quantityj ¼ αpricej
This demand curve is shaped as shown in Figure 1. The curvature and placement of the demand curve are determined by the parameters α and β, where β is negative. If we take natural logarithms of this equation (a simple transformation) the resulting relationship is linear and therefore easy to estimate using the same methods as in the simple linear model:
520
introduction
521
Price
Quantity
Appendix B – Figure 1
Non-linear demand curve
Inðquantityj Þ ¼ InðαÞ + βIn ðpricej Þ As a technical point, the use of natural logarithms means that the coefficient β can be interpreted as the price elasticity of demand – as such this demand curve exhibits constant elasticity at any price. The constant elasticity demand curve is just one form of non-linear demand function that can be empirically identified. However, while the assumption of constant elasticity allows a demand curve to be identified relatively easily, the assumption of this functional form imposes restrictions on the shape of the demand curve. The challenge in empirical analysis is to find a modelling framework that balances ease of computation (which generally requires the imposition of more assumptions regarding consumer behaviour) and flexibility (for which fewer assumptions are preferable). A popular demand model frequently used in empirical research is the Almost Ideal Demand System (AIDS).1 The AIDS model makes use of well-grounded assumptions from individual level microeconomic theory: namely, that consumers have a utility function that they maximise subject to a budget constraint. This is an important assumption that allows us to use price and quantity data, in addition to individuals’ expenditure, to derive a demand function. A convenient property of the AIDS model is that if 1
A. Deaton and J. Muellbauer, (1980) 70(3) The American Economic Review 312–26.
522
models for demand estimation
all individuals in a given multi-product market conform to the restrictions imposed by the demand equation then so does the aggregate. The AIDS model yields a demand curve that provides numerous parameters of interest, including own- and cross-price elasticities, from a model with strong foundations in fundamental microeconomic theory. Nonetheless, while the AIDS model is relatively parsimonious in its data requirements, the number of parameters to be estimated increases rapidly with the number of products analysed. In a market featuring 50 differentiated products, for instance, it would be necessary to estimate 502, that is, 2,500 different parameters.
1.2 Models for discrete demand estimation The AIDS model is based on the assumption that consumers make purchasing decisions in order to maximise some utility function. This function is likely, in its most complete form, to include prices and other product characteristics that are both observable and unobservable to the econometrician. Discrete choice analysis is used to characterise the choice made by a consumer when choosing a single product from a number of options. For example, when considering how to travel from London to Brussels a consumer would not be considering how much of a particular mode of transport she should buy given prevailing prices, but which mode of transport she should buy given prevailing prices. When she chooses to travel by aeroplane she makes a discrete choice that, in theory, has maximised her utility subject to a budget constraint. A fundamental assumption is that agents choose the alternative which provides them with the highest utility. The utility derived from a particular choice is often assumed to be a linear function of a set of choice-specific observable variables, a set of individual-specific observable variables and an unobservable random error term. For example, a consumer’s utility from choosing a particular car may be related to a number of car-specific variables (size, fuel efficiency, etc.), a number of variables specific to the individual (number of family members, income, age) and some components that we cannot observe. A complication of this approach is that we must make assumptions about what we cannot observe in the model. The assumptions we choose to make will affect the results of our analysis. A popular assumption for discrete choice analysis is that the unobserved variable follows a ‘type 1 extreme value distribution’ (the details of which are not important for this brief treatment), which gives the conditional logit demand
introduction
523
model.2 This model provides, for each good, the conditional probability that an individual selects that good above all others. If one has data on individual characteristics such as income or number of family members then these data can be taken into account to produce an estimate of the probability that an individual selects a particular good above all others given that individual’s characteristics. Multiplication of these probabilities by the total number of consumers in the market therefore describes the total quantity demanded for each product. Often, particularly during merger simulation, we are interested in the odds-ratio. This is the probability that one good, i, is chosen relative to the probability that another good, j, is chosen. Thus, for example, if the odds-ratio between i and j is 2, this implies that a new consumer entering the market is twice as likely to select good i than j. The assumptions of the conditional logit model allow the odds-ratio between products to be readily calculated in terms of observed variables. However, these assumptions also imply that the odds-ratio between goods i and j is independent of (a) any good other than i or j, and (b) the number of alternative products in the market. This property is termed the independence of irrelevant alternatives (IIA). IIA implies that the probability that one good is chosen relative to the probability that another good is chosen does not change if a third alternative is added (or has its characteristics changed). Unfortunately this attribute is counterintuitive in many cases. For example, consider the following classic example (from McFadden in 1973) of a commuter’s transport market, in which there are two possible methods of getting to work: by red bus or by car.3 Suppose that a consumer chooses each of these options with probability 1/2 (implying an odds-ratio of 1). Suppose now that another option, a blue bus, is added to the set of alternative means of transport and that consumers that choose to travel by bus are completely indifferent as regards its colour (red buses and blue buses are perfect substitutes). Intuitively, one might expect that consumers would still choose ‘car’ with probability 1/2, and ‘bus’ with probability 1/2, implying that red bus and blue bus are each chosen with probability of
2
3
See D. McFadden, ‘Conditional Logit Analysis of Qualitative Choice Behaviour’, in P. Zaermbka (ed.), Frontiers in Econometrics (New York: Academic Press, 1973). McFadden was not the first to point out the problem with IIA. This was Gérard Debreu in 1960, who used the slightly more highbrow example of the market for the music of Ludwig van Beethoven and Claude Debussy.
524
models for demand estimation
Bus
Red Bus
Car
Blue Bus
Appendix B – Figure 2 Structure of consumer’s decision-making process
1/4. However, this would violate IIA. This is because, if IIA holds, the odds-ratio for car/red bus (previously 1) should not be changed following the addition of blue bus as an additional alternative. Indeed, in this situation IIA would require that car, red bus and blue bus are each chosen with probability 1/3. IIA fails to take into account that red bus and blue bus are closer substitutes to one another than either is to car. The IIA assumption is the source of the criticism that logit demand models assume that switching to alternative brands is in proportion to each brand’s share of supply amongst the set of alternatives. Common sense tells us that switching from a given brand need not be in proportion to brand share and that certain brands may be ‘closer’ to one another. The conditional logit method can be adjusted to account for this problem via the use of nests, which cluster goods together based upon their attributes. In the above example, for instance, we might place the red bus and blue bus together in the same nest, such that the first decision made by the commuter is bus versus car, and the second decision (where bus is chosen) is red bus versus blue bus. The structure of the consumer’s decision-making process in this framework is represented graphically in Figure 2. This method of estimating a demand framework is known as the nested logit model. It is essentially a logit as described above applied to each stage of the decision process. It enables researchers to specify groups of alternatives that are similar to each other (allowing for some
introduction
525
correlation of consumer tastes across products). By grouping alternatives within a particular nest into sub-nests, this can be generalised so that there are multiple nesting levels. The nested logit model is often used in economic research and was recently used in Kraft Foods/Cadbury (put forward in evidence by the parties as a part of their merger simulation) and in Unilever/Sara Lee Body Care. As always this estimation method has its pros and cons. One clear advantage it has is that the number of parameters to be estimated is vastly smaller than with the AIDS model described above. However, this comes at a cost. The (relative) ease with which the model is estimated is a consequence of the rigid structure on substitution patterns imposed by the choice of nests and the allocation of products between those nests. These choices, which are crucial in determining the outcome of the model, must be made a priori by the econometrician on the basis of other evidence or intuition. Consequently, to a large extent, the outcome of nested logit models may derive not from the econometric analysis of the data but from somewhat arbitrary decisions made by the researcher. As such, the nested logit model’s inability to provide objective guidance on substitution patterns between products may in some cases make it an unattractive approach in the context of merger assessment. For a more detailed discussion of discrete choice and demand estimation more generally, see, for example, Davis and Garcés.4 4
P. Davis and E. Garcés, Quantitative Techniques for Competition and Antitrust Analysis (Princeton University Press, 2009).
INDEX
sponsoring new entry 296–7 conclusions 297 and coordinated effects, assessing see external stability condition relevance in EU cases 290–2 Conner effect 291 price discrimination/ differentiation and smaller customers 291–2 vulnerability of smaller customers 291
anti-competitive foreclosure 378–9, 386–7, 389 Antitrust Logit Model (ALM) 216 asymmetry and collusion 330–2, 339 capacity, asymmetry in 331–2, 371 costs, asymmetry in 331, 371 increasing symmetry 371–2 multi-market contact/asymmetry in individual market 360 substitutable product varieties, asymmetry in number of 331, 371 benchmarking correlation 111–14, 129–30 Bertrand model 151, 153–4, 216 and Cournot competition 153 merger simulation 227–30 ‘binary fallacy’ 37–8 bundling 387 buyer power 288–97 buyer power in EC Horizontal Merger Guidelines 288–90 ability to prevent supplier from increasing prices 289, 292–7 buyer power created by mergers 289–90 buyers must be of considerable size 289, 291 definition of countervailing buyer power 289 nature of buyer power 288–9 case study: the Enso/Stora decision 292–7 small losses in volume significantly impacting suppliers’ average costs 293–5
capacity collusion 330–2, 351 asymmetry in capacity 331–2 excess industry capacity 333–4 retaliation and excess capacity 362 rivals rivals’ ability to expand output/ ease of adding capacity 261–4 rivals’ capacity in EC Horizontal Merger Guidelines 258–9 ‘cellophane fallacy’ 38–9 chains of substitution 41–3 Chicago School 388 Chief Economist Team/Chief Competition Economist, EU 12–13 benchmarking correlation 112 shock analysis/supply outages 136 collusive behaviour see horizontal mergers: coordinated effects commitments/remedies 8–10 common costs 107–11, 129 Community dimension see under concentrations
526
index competition and elasticity 26–7 number of competitors see number of competitors and collusion policy 24–5 unilateral effects see horizontal mergers: unilateral effects competitive constraints see important competitive constraints; important competitive competitive fringe see under external stability condition complementary and substitutable products 380–2 concentrations application of EC Merger Regulation 3–4 Community dimension 4 mandatory notification 7 definition 3 see also legal framework and Commission procedure conglomerate mergers 2, 378 anti-competitive foreclosure 387 economies of scope, giving rise to 384 ‘one monopoly profit’ 388 coordinated effects see horizontal mergers: coordinated effects cost curve 20–2 economies of scale 21–2 fixed and variable costs 20 marginal cost 20–2 sunk costs 22 countervailing factors 150, 288–317 buyer power see buyer power efficiency analysis see efficiency analysis failing firm defence see failing firm defence product re-positioning and new entry see product re-positioning and new entry Cournot, Antoine 93 Cournot model 150–3 and Bertrand competition 153 Cournot effect 382 market price decreasing as number of competing firms increases 151
527
predicted prices exhibiting direct relationship with HHI 152–3 critical discount factor see under economic concepts of tacit coordination critical loss analysis 46–64 description of technique 46–57 calculation of the critical loss in practice 52–6 data requirements 56–7 SSNIP and critical loss analysis 51–2 SSNIP and pricing decisions 46–51 example 58–9 use in EU merger control 59–64 Commission undertaking own critical loss analysis 63–4 critical loss analysis not appropriate in airline industry 59–62 customer switching 35, 76–7, 94–5 customer switching analysis see customer switching analysis 156 differentiated products 156 greater switching propensity and greater unilateral effects 148 switching costs see switching costs customer switching analysis 183–91 description of technique 183–5 aim/focus of switching analysis 183–4 overlapping customers analysis 184–5, 190 example 185–9 use in EU merger control 189–91 assessing degree of closeness of competition 189–91 customers buyer power see buyer power customer/market sharing 338, 340–1, 374–5 infra-marginal customers 35–6, 76–7 marginal customers 35, 76–7, 94–5 overlapping customers 184–5
528
index
customers (cont.) switching see customer switching, see also survey evidence: market definition; survey evidence: competitive constraints data room process for third party information 9 Davis, P.J. 371–2 demand curve 14–20 elasticity of demand see elasticity of demand marginal revenue 17–18 demand estimation 28, 64–74 description of technique 64–9 demand estimation and market definition 69 empirical demand estimation and identification 65–9 example 69–71 use in EU merger control 71–4 demand elasticity estimation in merger analysis seldom used 71–2 unsafe to rely solely on demand estimation analysis 73–4 demand fluctuations/lumpiness of demand and collusion 329–30, 367 demand growth and collusion 329 demand-side substitution 34–5, 94–5, 136 diagonal mergers 390 Dickey-Fuller test 129–31 differentiated products/services closeness of competition and differentiated products 161 competitive constraints in industries with differentiated products 148–9 coordination terms agreement more difficult 338–40 definition of relevant market in assessing unilateral effects 148–9 diversion ratios see diversion ratios HHIs not taking account of product differentiation 286
merger simulation models 215 mergers in markets where goods/services are differentiated 148–9 price levels analysis 95–6 product specifications differentiation as evidence of separate national markets 103–4 diversion ratios 156 approximating 184 calculated from responses to hypothetical price increases 179–80 combined with survey information on merging firms’ gross margins 183 example of diversion ratios in merger assessment 157–8 price pressure tests 218 use of surveys to estimate diversion ratios 177 dominance see monopoly or dominant positions double marginalisation, elimination of 382 EC Horizontal Merger Guidelines see horizontal mergers: unilateral effects econometric analyses/techniques 43, 72–4, 129–30, 133 bidding data analyses 242, 245–6 case study: Statoil/JET 203–6 econometric estimates 204–5 JET’s role 203–4 placing econometrics in right economic context 205–6 complex econometric analyses remaining the exception 27–8 demand estimation 65, 69–71 economic evidence provided by structural econometric models 29 entry analysis 209–10 instrumental variables 68–9 as key source of quantitative evidence on market definition 65
index nature of 3 price/concentration studies 192, 198–206 ‘reduced form’ econometric models 29 scanner data for mergers in branded consumer goods industries 215 economic concepts, basic 3, 13–27 cost curve see cost curve demand curve see demand curve market power see market power profit maximisation see profit maximisation economic concepts of non-horizontal mergers 379–91 foreclosure and anti-competitive foreclosure 385–9 anti-competitive foreclosure 378–9, 386–7, 389 foreclosure 386 ‘one monopoly profit’ 388–9 other anti-competitive effects 389–91 coordinated effects 390–1 other non-coordinated effects 389–90 pro-competitive effects 380–5 complementary and substitutable products 380–2 economies of scope, conglomerate mergers giving rise to 384 ‘efficiency offence’ 385 ‘hold-up’ problem, nonhorizontal mergers eliminating 383–4 merger between complementary products giving price reductions 382 supply chain efficiencies, vertical merger generating 384 supply side efficiencies 383–5 economic concepts of tacit coordination 319–35, 369 factors that affect the critical discount factor 326–35 asymmetry (+) 330–2 demand fluctuations/lumpiness of demand (+) 329–30, 367 demand growth ( ) 329
529
ease of entry (+) 328 excess industry capacity (?) 333–4 frequency of interaction/price adjustments ( ) 328 innovation (+) 332–3 market demand elasticity (?) 334–5 multi-market contact ( ) 332 number of competitors (+) 327–8 substitutability, degree of (?) 333 transparency ( ) 328–9 textbook tacit coordination 321–6 coordination sustainable if firms sufficiently ‘patient’ 324–5 critical discount factor 323–5 how firms reach collusive agreement, repeated games theory silent on 325–6 modification to textbook model allowing tacit coordination 323–5 no framework to identify change from competitive to collusive 326, 337 non-cooperative game theory/ repeated games theory 321 ‘textbook’ model of coordination 321–3 theory silent on how firms select between equilibria 369 economic theory: important competitive constraints 150–8 closeness of competition and diversion ratios 155–6 example of diversion ratios in merger assessment 157–8 economic models and market concentration 150–4 Bertrand model 151, 153–4 Cournot model 150–3 economics in EU merger control, role of 2, 12–13 Chief Competition Economist post see Chief Competition Economist, EU ‘effects-based’ approach, consequences of 12
530
index
economies of scale 21–2 economies of scope 384 ‘effects-based’ and ‘form-based’ approaches 12 efficiency analysis 304–10 efficiency analysis in EC Horizontal Merger Guidelines 304–7 consumer benefit 305–6 merger specificity 305 requirements to sustain efficiency defence 305–6 sufficiency of claimed efficiencies to offset lessening of competition 307 verifiability 306 relevance in EU cases 308–10 elasticity of demand 16–20, 80 and competition 26–7 inelastic demand as sufficient to identify a relevant market 73 market demand elasticity and collusion 334–5 and market power 25–6 Elzinga, Kenneth 81 see also LIFO/ LOFI (Elzinga-Hogarty) test empirical economic evidence 3, 27–9 economic evidence provided by structural econometric models 29 empirical evidence comprising basic statistics 28–9 empirical evidence deriving from use of ‘reduced form’ 29 methods unified by use of statistical inference 29 empirical techniques: competitive constraints see important competitive constraints between the parties: empirical techniques empirical techniques to assess market definition 43–126 analysis of price levels see price levels analysis analysis of sales patterns see sales patterns analysis critical loss analysis see critical loss analysis
demand estimation see demand estimation price correlation analysis see price correlation analysis qualitative evidence see qualitative evidence shock analysis see shock analysis stationarity analysis see stationarity analysis survey evidence see survey evidence: market definition entry analysis 207–10 description of technique 207 example 207–8 use in EU cases 209–10 entry of new suppliers or products entry analysis see entry analysis product re-positioning and new entry see product re-positioning and new entry shock analysis 134–5 sponsoring new entry/buyer power 296–7, 367 ease of entry and collusion 328 Epstein, R. 216 equality of prices see under price levels analysis exchange rate shocks or fluctuations see under shock analysis external stability condition 319, 349–50, 362–8 buyer power 367–8 assessment of role of buyers in Pirelli/BICC 367–8 bulk purchases influencing demand fluctuations/lumpiness of demand 367 sponsoring new entry/growth of smaller players 367 undermining stability of coordination 367 importance of buyer power and potential competition 363 non-colluding rivals 363–7 assessment of role of fringe in ABF/GBI Business 366–7
index assessment of role of fringe in Airtours/First Choice 363–4 assessment of role of fringe in Norske Skog/Parenco/Walsum and UPM-Kymmene/Haindl 365–6 colluding firms constrained by expansion of competitive fringe 363, 367 non-horizontal mergers 391 failing firm defence 310–17 failing firm defence in EC Horizontal Merger Guidelines 310–12 conditions for use of failing firm defence 311–13 high hurdle for failing firm defence 311–13, 317 nature of failing firm defence 310–11 purposes of tests 312 relevance in EU cases 312–17 cases providing the foundation for the three criteria 313–15 ‘failing division’ argument 316–17 rejection of failing firm defence in recent investigations 315–16 Farrell, J. 61–2, 220 fixed costs 20 efficiency analysis/reductions and pricing 305–6, 308–9 fixed cost spreading 21–2 Foncel, J. 217 foreclosure 2, 386 framework for the assessment of coordinated effects 320–1, 335–77 ability of firms to reach coordination terms, factors affecting 327, 338–9 complexity of market 338–9 institutional features of market facilitating coordination 339 number of players 338 supply and demand conditions, stability of 339 symmetry of firms 339
531
market characteristics so that tacit understanding likely to be sustained? 336–7, 349–68 external stability see external stability internal stability see internal stability merged entity/remaining competitors able to reach tacit understanding? 336–49 complex or non-transparent market conditions 338–41 customer/market sharing 338, 340–1, 374–5 innovation 338–9, 341 non-price parameters of competition, coordination on 336, 340 parameters over which firms likely to agree to coordinate 337–41 price coordination 338–41 theory of coordination in ABF/ GBI Business 346–9 theory of coordination in Sony/ BMG 341–5 more systematic approach to assessment of coordinated effects 336–7 past coordination, account taken of 337 lower threshold 337 proposed transaction makes tacit coordination likely or more effective? 337, 368–77 assessment of effect of merger in ABF/GBI Business 375–7 merger specific factors see merger specific factors affecting likelihood and effectiveness of tacit coordination when coordinated effects/concerns arise 368–9 traditional ‘checklist approach’ to assessment of coordinated effects 335–6 fringe/competitive fringe see under external stability condition Froeb, L. 216
532
index
game theory see economic concepts of tacit coordination Gross Upward Pricing Pressure Index (GUPPI) 219–22, 231 modified for use in partial ownership cases 286–7 harm, theory of 206 coordinated effects theory of harm 2, 12, 337 pre-existing coordination 337 threshold for harm lower if existing tacit coordination strengthened 337, 369 see also framework for the assessment of coordinated effects non- horizontal theory of harm 12 Herfindahl–Hirschman Index (HHI) 152–3 modified to take account of incentives changes from minority shareholdings 285–6 limited value 286 non-binding presumptions based on HHI 160 Hogarty, Thomas 81 see also LIFO/ LOFI (Elzinga-Hogarty) test ‘hold-up’ problem 383–4 horizontal mergers: coordinated effects 318–77 coordinated effects, nature of 318 and unilateral effects 320, 326 framework for assessment see framework for the assessment of coordinated effects tacit coordination collusive outcomes easier to achieve/sustain in concentrated markets 319 economic concept see economic concept of tacit coordination elements required to arise/be sustained 318–19 externally sustainable collusive agreement see external stability condition
internally sustainable collusive agreement see internal stability condition mergers increasing market concentration increasing the risk of 319 nature of 318 horizontal mergers: unilateral effects 148–295, 380 coordinated effects and unilateral effects 320, 326 countervailing factors see countervailing factors economic concept of unilateral effects analogous to single firm dominance 148–9 further issues relevant in the assessment of unilateral effects 149–50, 251–87 case study: analysis of potential entry in UK Svitzer/Adsteam case 255–8 elimination of potential competition in EC Horizontal Merger Guidelines 252–8 elimination of potential competition in EU precedents 253–5 partial ownership see partial ownership rivals’ ability to increase supply see rivals’ ability to increase supply switching costs see switching costs important competitive constraints see important competitive constraints; important competitive constraints between the parties: empirical techniques when mergers give rise to unilateral effects 148–9 Huse, C. 371–2 hypothetical monopolist/‘SSNIP’ test see under market definition identification 67–9 Illustrative Price Rise (IPR) 218–19, 231
index import duties, changes in see under shock analysis important competitive constraints 148–76 Commission’s recent enforcement practice 163–76 differentiated products see under differentiated products/services EC Horizontal Merger Guidelines see important competitive constraints: EC Horizontal Merger Guidelines economic theory see economic theory: important competitive constraints strength of competitive constraints proportionate to market share 148 when important competitive constraints likely to exist 158–61 important competitive constraints: EC Horizontal Merger Guidelines 158–63 merger eliminates an ‘important competitive force’ 149, 159, 162–3 ‘maverick firms’ 162–3, 165 merging firms are close competitors 160–2, 165, 177 degree of substitutability 161, 165 differentiated products and closeness of competition 161 likelihood of unilateral effects inferred from market shares 160–1 remaining rivals not providing effective competitive constraint 161–2 merging firms have large market share 159–61, 165, 176 market shares and scope for unilateral price increases 159 market shares thresholds 159–60, 176 non binding presumptions based on HHI 160
533
‘safe harbour’ for firms with market share below 25% 159, 176 unilateral effect concerns if combined market shares over 50% 159–60, 165, 176 important competitive constraints between the parties: empirical techniques 149, 176–251 customer switching analysis see customer switching analysis entry analysis see entry analysis merger simulation see merger simulation natural experiments see natural experiments price/concentration studies see price/ concentration studies and analysis of the impact of rivals’ presence survey evidence see survey evidence: competitive constraints win/loss and bidding analysis see win/loss and bidding analysis ‘important competitive force’ see under important competitive constraints: EC Horizontal Merger Guidelines industrial organisation 13–14, 25–6 infra-marginal customers see under customers innovation and collusion 332–3, 338–9, 341 instrumental variables 68–9 internal stability condition 319, 349–62 monitoring deviations 350–5 assessing degree of transparency/firms inferring rivals’ actions 350 assessing degree of transparency/ firms’ information on rivals’ decisions 351 reducing the number of market participants, effect of 371 transparency affected by way transactions take place 350 transparency higher when fewer firms active in market 350
534
index
internal stability condition (cont.) monitoring in ABF/GBI Business 351–2 monitoring in Sony/BMG (2007) 351–5 retaliation see retaliation isoelastic demand 56 Ivaldi, M. 217 lagged responses 114–15 legal framework and Commission procedure 2–11 EC Merger Regulation 3–7 addressing enforcement gap in 1989 Merger Regulation 5–7 ‘concentrations’, application to 3–4 European Commission jurisdiction 4 permitted mergers 3–4 substantial impediment to effective competition/ competitive assessment 4–6 procedure for notification and assessment 7–11 Art. 6 approval/conditional approval decisions 9 Art. 8 clearance/clearance subject to commitments/prohibition 9–10 doubts triggering Phase II investigation 9–10 mandatory notification and Phase I competitive assessment 7–9 Lerner condition 61 LIFO/LOFI (Elzinga-Hogarty) test 81–5 linear demand 56 marginal cost 20–2 profit maximisation 23–4 marginal customers see under customers marginal revenue 17–18 profit maximisation 23–4 market concentration and economic models see economic theory and key concepts: competitive constraints
market definition 31–126 asymmetry in market definition 39–40 conceptual framework 33–42 hypothetical monopolist/‘SSNIP’ test 32–6 issues and complications arising in market definition 36–42 empirical techniques see empirical techniques to assess market definition indirect constraints and chains of substitution 41–3 Market Definition Notice/ identifying relevant market 31–2, 43, 45–6, 54, 65, 75, 105 ‘SSNIP’/hypothetical monopolist test 32–6, 39–40, 43, 94–5 critical loss analysis see under critical loss analysis magnitude of SSNIP and the ‘binary fallacy’ 36–8 plant shutdowns 136 relevant pre-SSNIP price and ‘cellophane fallacy’ 38–9 market investigation 8–9, 75 market power 24–7 constraints on buyer power see buyer power potential competition see horizontal mergers: unilateral effects and demand elasticity 25–6 ‘one monopoly profit’ 388–9 market shares see under EC Horizontal Merger Guidelines: important competitive constraints market sharing 338, 340–1, 374–5 ‘maverick firms’ 162–3, 165 meaning of ‘maverick’ 373–4 mergers creating 374 Merger Regulation, EC see under legal framework and Commission procedure merger simulation 214–31
index description of technique 214–23 as highly technical/data intensive process 214–15 methodological challenges 215 ‘nests’ in demand assumptions, introducing 217 predictions dependent on key technical parameters 217–18 price pressure tests see price pressure tests principle of merger simulation 214 simple simulation models, development of 216 simplifying assumptions reflecting central elements of competition 215–16 statistical estimates of individual own/cross-price elasticities as inputs 217 whether models provide reliable guide to actual effects of mergers 218 example 223–6 use in EU cases 226–31 Commission’s general support for technique 228–31 restrictive assumptions/use in relatively limited number of cases 226–7 merger specific factors affecting likelihood and effectiveness of tacit coordination 370–5 increasing market segmentation and retaliation possibilities 374–5 increasing symmetry 371–2 reducing the number of market participants 370–1 removing a ‘maverick’ 372–4 monopoly or dominant positions 165–76 dominance threshold 176 ‘one monopoly profit’ 388–9 single firm dominance 148–9, 165 tacit coordination as collective dominance 319–20 assessing see framework for the assessment of coordinated effects
535
Moresi, S. 221 multi-market contact and collusion 332, 359–62, 374–5 non-horizontal mergers 391 ‘multi-sourcing’ 84–5, 291 Nash Equilibrium 323 natural experiments 211–14 description of technique 211 example 211–13 rivals’ ability and incentive to increase output 266–7 use in EU cases 213–14 ‘nests’ in demand assumptions 217 nested-logit demand 227–30 non-colluding rivals see under external stability condition non-cooperative game theory see economic concepts of tacit coordination non-coordinated effects see horizontal mergers: unilateral effects non- horizontal mergers 2, 378–475 economic concepts see economic concepts of non-horizontal mergers less likely to be anti-competitive 378–9 meaning 378 non- horizontal theory of harm 12 notification and assessment see under legal framework and Commission procedure number of competitors and collusion 327–8, 338 non-horizontal mergers 390–1 reducing the number of market participants 370–1, 390–1 O’Brien, D. 286–7 OECD Policy Roundtables paper on minority shareholdings 286 oligopoly theory: theoretical models see Bertrand model; Cournot model
536
index
partial ownership 282–7 example of effects on competitive behaviour/market outcomes 283–7 modified HHI 285 increasingly prominent issue 282–3 minority shareholdings in absence of control 282 plant outages or strikes see shock analysis potential competition constraint on market power see horizontal mergers: unilateral effects and coordinated effects, assessing see external stability condition price/concentration studies and analysis of the impact of rivals’ presence 191–207 case study: Statoil/JET see under econometric analyses/ techniques description of technique 191–2 markets differing in respects other than the level of concentration 192 selection of suitable comparator markets as most important element 191–2 example 192–201 use in EU merger control 201–3 price correlation analysis 105–27, 134, 137 description of technique 105–15 benchmarking correlation 111–14, 129–30 lagged responses 114–15 principle on which price correlation analysis based 105–6 spurious correlation 107–11, 129 example 115–19 use in EU merger control 119–27 downside of correlation analysis 126 evidence to be viewed in context of other forms of evidence 119
price-cost margin 25–6 price elasticity high margins and low price elasticity 60–1 merger simulation 217 surveys not capturing customers’ average willingness to switch 80 price levels analysis 93–105 description of technique 93–7 differentiated products 95–6 equality of prices 93–5 geographic market definition 96–7 example 97–100 use in EU merger control 100–5 Commission using analysis of price levels in defining relevant markets 103 recent citing of price level evidence 100 unsafe to draw conclusions on market definition based on price levels 103 price pressure tests 218–23 aim to provide initial screen not in-depth investigation 222 criticised 222–3 unifying feature 218 variants of price pressure tests 218–22 Gross Upward Pricing Pressure Index (GUPPI) 219–22, 231 Illustrative Price Rise (IPR) 218–19, 231 Upward Pricing Pressure (UPP) 219–21, 231 pricing decisions coordination on price 338–41 differentiated products 339–40 frequency of interaction/price adjustments and collusion 328 Cournot effect/elimination of double marginalisation 382 incentives to increase price 389–91 and SSNIP 46–51 product re-positioning and new entry 298–304
index in EC Horizontal Merger Guidelines 298–302 brand re-positioning changing product characteristics or marketing 299 circumstances where threat of entry more likely to be credible 302 markets with large buyers and significant sunk costs 301–2 need to understand how other firms would respond to increased prices 300 new entry to market reducing concerns over unilateral price increases 299 new entry more effective if done relatively quickly and cheaply 300–1 product re-positioning as effective countervailing factor 298–9 product re-positioning reducing incentive for unilateral price increases 298 relevant benchmark for assessing viability of entry 299–300 re-positioning by non-merging entities 298 threat of exercising buyer power constraining post-merger prices 302 relevance in EU cases 302–4 entry barriers 303–4 mere possibility of entry 304 supply-side response arguments 302–3 profit maximisation 15, 23–4 firms passing on part of variable cost reduction to customers by lower prices 304–5 fixed cost spreading 21–2 organising to set marginal revenue equal to marginal cost 23–4 Proportionally Calibrated Almost Ideal Demand System (PCAIDS) model 216
537
qualitative evidence 27–8, 44–5, 78, 119, 180 forms in market definition analysis 44 no hierarchy between ‘technical’ and ‘non-technical’ evidence 9–10 survey information as qualitative information on customer preferences 177 quantitative evidence see empirical techniques to assess market definition regression analysis 28, 66, 103–4 price/concentration studies 198–201 repeated game theory see economic concepts of tacit coordination retaliation 355–62, 374–5 credible punishment may not be effective 357–8 excess capacity 362 multi-market contact 359–62 timing of retaliation 358–9 effective punishment may not be credible 356–7 Norske Skog/Parenco/Walsum/ UPM-Kymmene/Haindl assessment 356–7 importance of establishing firms can punish deviation 355–6 rivals ability to increase supply see rivals’ ability to increase supply anti-competitive foreclosure 378–9, 386–7, 389 firms’ information on rivals’ decisions 351 foreclosure 386 inferring rivals’ actions 350 harmed by mergers/‘efficiency offence’ 385 non-colluding see external stability condition rivals’ ability to increase supply 258–75
538
index
rivals’ ability to increase supply (cont.) assessing mergers in electricity markets 267–75 case study: Gas Natural/Endesa 273–5 market characteristics 268–70 unilateral effect concerns 270–3 EU cases 261–7 rivals’ ability to expand output/ ease of adding capacity 261–4 rivals’ incentive to expand output 264–7 example 259–61 rivals’ capacity in EC Horizontal Merger Guidelines 258–9 Rubinfeld, D. 216 sales patterns analysis 81–93 description of technique 81–5 geographic market definition 82–3 product market definition – ‘multi-sourcing’ 84–5 example 85–8 use in EU merger control 88–93 potential distinction between transport modes 92–3 potential impact of shipment size 93 proportion of sales to be considered/threshold 91–2 Salop, S.C. 221, 286–7 Shapiro, C. 61–2, 220 shock analysis 134–47 description of technique 134–8 changes in exchange rates/ import duties, asymmetric pricing from 137–8 entry of new suppliers or products 134–5 nature of shock analysis 134 supply outages/plant shutdowns affecting firm’s supply ability 135–7 example 138–44 use in EU merger control 144–7
entry 144–5 exchange rate shocks or fluctuations 147 plant outages or strikes 145–7 single firm dominance see monopoly or dominant positions spurious correlation 107–11, 129 ‘SSNIP’ test see market definition statement of objections 10 stationarity analysis 105, 127–34 description of technique 127–30 advantages as alternative to price correlation 129–30 increasingly important role of stationarity analysis 127 rationale 128 example 130–1 use in EU merger control 131–4 more sceptical view of stationarity analysis taken 132–4 price stationarity not dispositive of competitive interaction 133 role of stationarity analysis within market definition 133–4 spurious stationarity 133 stationarity analysis as complement to correlation analysis 131–2 strikes or plant outages see under shock analysis substitutability between products complementary and substitutable products 380–2 degree of substitutability and collusion 333 unilateral effects see horizontal mergers: unilateral effects sunk costs 22 product re-positioning and new entry 301–2 supply chain efficiencies, vertical merger generating 384 supply outages natural experiments 211 shock analysis see shock analysis supply side efficiencies see economic concepts of non-horizontal mergers
index supply-side response arguments 302–3 supply-side substitution 34, 136 survey evidence: competitive constraints 177–83 description of technique 177–9 customers identifying range of credible alternative suppliers 178 customers ranking suppliers for products’ technical features/ pricing 179 diversion ratios from responses to hypothetical price increases 179–80 important role of customer surveys in merger assessment process 178 need to adhere to principles of survey design 177–8 survey information as qualitative information on customer preferences 177 surveys as source of quantitative information on degree of substitution 178 example 179–80 use in EU merger control 180–3 diversion ratios evidence and survey information on gross margins 183 use of large-scale consumer surveys 180–1 survey evidence: market definition 74–80 description of technique 74–8 customer surveys 75–8 market investigation see market investigation problems with customer surveys 76–8, 177 use in merger control 78–80 Commission substituting own views for consumers’/evidence value 80 qualitative evidence 78–9 surveys not capturing customers’ average willingness to switch 80
539
switching see customer switching; switching costs switching costs 275–82 conclusions 281–2 effects of switching costs on competition 276–7 implications of switching costs for merger assessment 277–81 Lloyds/Abbey National 278–9 Lloyds/HBOS 280–1 tacit coordination see under horizontal mergers: coordinated effects transparency and collusion 328–9 complex or non-transparent market conditions 341 monitoring deviations 350–5 assessing degree of transparency/ firms inferring rivals’ actions 350 assessing degree of transparency/ firms’ information on rivals’ decisions 351 transparency affected by way transactions take place 350 transparency higher when fewer firms active in market 350 non-horizontal mergers 391 timing of retaliation/time lag 358 tying 387 undertakings 165, 268 unilateral effects see horizontal mergers: unilateral effects Upward Pricing Pressure (UPP) 219–21, 231 variable costs 20 efficiency analysis/reductions and pricing 305–6, 309 Verboven, F. 217 vertical mergers 2, 378 anti-competitive foreclosure 387 ‘one monopoly profit’ 388 Werden, G. 216 win/loss and bidding analysis 231–51 case study: Pirelli/BICC 248–51 analysis of credible bidders 251
540
index
win/loss and bidding analysis (cont.) bidding competition 250 transitional state of the industry 249–50 description of technique 231–6 analysis of the relationship between prices and number of bidders 235–6 analysis of the relationship between prices and runner-up/ bidders 235 competition between suppliers by bidding for customer contracts 232–4 participation analysis 234–5 runner-up analysis 235 example 236–41 analysis of the relationship between prices and identity of runner-up 239–40
analysis of the relationship between prices and number of bidders 240–1 participation analysis 236–9 runner-up analysis 239 use in EU merger control 241–8 comprehensive and representative bid data samples 246–7 most frequent reliance on participation analyses 242–4 overstating importance of less competitively significant firms 244–5 runner-up analyses, advantages of 244–5 segments representing a separate market 246